Juji - the easist DIY AI chatbot platform

Transforming Academic Advising with AI Technology

2024-07-15T14:30:56Z

In the rapidly evolving landscape of higher education, academic advising stands as a critical component in fostering student success. However, advisors often face numerous challenges, from managing large caseloads to providing personalized support. This blog delves into the transformative potential of AI in this domain. Below we first discuss three key challenges in academic advising and how AI can provide innovative solutions. Then we offer a practical guide on applying the NACADA Academic Advising Core Competencies Model to building effective AI Advising Assistants. Join us as we uncover how AI can revolutionize academic advising, making it more efficient, personalized, and impactful for students and advisors alike.

3 Challenges in Academic Advising and How AI Can Help

Academic advisors play a crucial role in guiding students through their educational journey, providing support in course selection, academic policies, and career opportunities, etc. With limited staffing resources relative to the growing number of students requiring individualized guidance, advisors often manage large caseloads, which leads to several challenges. Below we highlight the key challenges and outline how the adoption of AI can effectively address them.

Challenge 1: Quick, Transactional Conversations

Due to the large number of students, advisors often have to keep their meetings brief and focused on immediate issues, resulting in conversations that are more transactional rather than transformational.

AI Solution: Scale Up Academic Advising

AI can provide advice to students 24/7, allowing advisors to scale their services and increase accessibility. This enables advisors to allocate more time for in-depth, transformational conversations with students, enhancing the advising experience.

Challenge 2: Limited Personalization

The large number of students and limited resources make it difficult for advisors to provide personalized attention, affecting the quality of advice and support.

AI Solution: Offer Personalized Support

AI can offer personalized information on majors, associated careers, required skills, and job demands, and provide tailored recommendations based on individual needs and preferences. This allows advisors to focus on more complex and nuanced student issues and engage in deeper, more meaningful conversations with students.

Challenge 3: Efficiency Over Depth

Advisors often need to prioritize efficiency, focusing on quick solutions rather than exploring complex issues in depth.

AI Solution: Support Data-Informed Decisions

AI can analyze student data to provide advisors with a "cheat sheet" and help them identify areas where students need additional guidance, enabling more targeted, in-depth support.

How to Apply NACADA Academic Advising Core Competencies Model to Building AI Advising Assistants

The National Academic Advising Association (NACADA), the premier association for student success through academic advising in higher education, has developed the Academic Advising Core Competencies Model to help the professional development of academic advisors. As we look to the future of using AI to augment academic advising, the same framework can also be used to guide the design, implementation, and evaluation of AI Advising Assistants.

Below we highlight how to apply each component of this model to AI Advising Assistants.

1. Conceptual Component to Guide AI Conversation Flow (Workflow)

The conceptual component covers the ideas and theories that human advisors should use to effectively advise their students. These theories and approaches provide guidance for structuring the conversation of AI Advising Assistants, enabling them to emulate effective advising practices.

2. Informational Component to Build AI Knowledge Base

The informational component encompasses the knowledge that human advisors must acquire to guide students at their institution. This includes institution-specific history, mission, vision, values, curriculum, degree programs, policies, procedures, and campus resources. Feeding this body of knowledge to AI Advising Assistants allows them to provide informed and contextually accurate responses to student inquiries.

3. Relational Component to Guide Functionality Development and Evaluation

The relational component lists the skills that human advisors must have to deliver effective advising per the above two components. It provides a blueprint for developing and evaluating the functions of an AI Advising Assistant. For example, an AI Advising Assistant should build good relationships with students, and facilitate problem-solving, planning, and decision-making.

A Reference AI Advising Assitant

Juji Studio enables organizations like universities to build and operate custom proactive AI chatbots to automate high-touch human interactions like academic advising. Acting as assistants to human academic advisors, the chatbots help establish deeper and more meaningful connections between advisors and students. They reduce the workload of advisors by handling routine information requests, analyzing student data to understand individual strengths, needs, and preferences, and crafting personalized advice. Advisors can focus on developing more effective advising strategies, addressing complex situations, and guiding AI to ensure the quality of its content.

The images below illustrate a multi-turn conversation between an AI Advising Assistant built using Juji Studio and a student seeking help with their financial situation. Developed based on the Academic Advising Core Competencies Model, the AI Advising Assistant demonstrates its capability to build on previous interactions, actively listen to the student's concerns, and express empathy. The chat showcases how the AI Advising Assistant follows a structured conversation flow to present multiple options to address the problem, provide clear and accurate information, and offer encouragement and collaboration with the student to tackle the issue together.

By aligning development with the NACADA Academic Advising Core Competencies Model, institutions can ensure that the AI can support advisors in achieving the advising objectives.

User-Centric Best Practices of Conversational AI Design

2024-07-10T17:16:52Z

Although powerful large language models (LLMs) like GPT can power a chatbot with abilities to interpret and generate response to user input, they do not really teach the chatbot how to carry out an effective and pleasant dialog with a user. Additionally, AI is far from perfect and there are always cases where AI cannot handle or handles wrong. Hence the role of a conversational AI designer is critical as they put their wisdom and creativity into an AI chatbot and ensure effective dialogs between AI and its users.

In the article “Towards User-Centric Guidelines for Chatbot Conversational Design,” published in the International Journal of Human-Computer Interaction, researchers Geovana Ramos Sousa Silva and Edna Dias Canedo at Universidade de Brasília conducted a systematic literature review to uncover best practices in conversational AI design around three areas:

Naturalness

A chatbot should engage in conversations naturally by:

Presenting itself as a virtual agent with a name and purpose at the start of the conversation.
Addressing the user by their name when possible.
Supporting small talk or chitchat, such as greetings and expressions of praise or thanks.
Echoing user input in responses when appropriate.
Conveying information in a conversational, casual tone.
Using alternative ways of saying the same thing, avoiding repetitive messages.

Emotionality

A chatbot should express and understand emotions to build deeper connections with users by:

Using punctuations, images, GIFs, and emojis to convey emotions and make conversations more engaging.
Showing empathy by adapting responses based on the user’s feelings.
Incorporating humor appropriately.

Transparency

A chatbot should be honest and clear about its capabilities to set the right user expectations by:

Presenting its purpose and capabilities at the start of the conversation.
Acknowledging its limitations upfront or after a failure.
Suggesting conversation topics to keep the user on track of what the chatbot can do.
Asking for clarification when it fails to understand user input.

In the following sections, we will share design tips on how conversational AI designers can apply these best practices when creating AI chatbots.

Best Practice Design Tips to Ensure Naturalness

To make AI-human conversations sound natural, a chatbot’s messages should incorporate elements commonly found in human-to-human interactions. Just like meeting any strangers, we often want to have a presence and leave a good first impression. The same applies to an AI chatbot design, giving your chatbot a suitable name along with an avatar (the presence), and even a distinct persona. A chatbot with an avatar (vs. a blank profile) makes users feel the existence or "realness" of the chatbot, helping the AI establish rapport and build trust with users and leading to more engaging interactions.

In Juji Studio, one can select from a variety of built-in avatars or upload their own to give a chatbot an identity. One can further customize the chatbot with a description of its personality (Image 1.1).

To exhibit the intended personality, a chatbot can use AI to auto-generate its messages based on its personality (Image 1.2). As always, human designers can easily customize or fine-tune AI-generated messages as they see fit.

Like in any conversations between strangers, a chatbot should also begin by greeting a user and introducing itself. As part of the introduction, it is always a good idea for the chatbot to explicitly address the user by their name as well as introduce its own name to kick off a conversation naturally (Image 1.3).

In human conversations, a good listener often echoes or paraphrases what their conversational partner says. This technique not only demonstrates that they are paying attention but also helps to build rapport and ensures mutual understanding. Similarly, in AI-human interactions, a chatbot should employ the same strategy to make the conversation feel more natural and engaging.

Juji Studio enables designers to instruct a chatbot to repeat what it hears in two ways. One way is to extract key information from user input and store it in a custom attribute, which can then be used in a chatbot response to acknowledge the received input or provide context for further conversations. Image 1.4 below demonstrates how to create a custom attribute from user input and use it to tailor a chatbot message.

Image 1.5 showcases this feature in action through an example conversation snippet.

The second way is to simply enable a chatbot to acknowledge what a user says by auto-generating its response, where Juji built-in prompt instructs the chatbot to repeat what it hears (Image 1.6).

In human conversations, when someone needs to repeat a message or question, they rarely use the exact same wording. Instead, they naturally vary their phrasing to keep the interaction dynamic and engaging, preventing it from sounding monotonous or robotic. Emulating this natural variability is crucial for AI chatbots to create more authentic and human-like interactions.

Juji Studio enables designers to define "paraphrases" for any chatbot message or request, either manually or automatically using GenAI (Image 1.7).

A unique feature of Juji Studio is that it allows designers to use the "only for re-asking" checkbox to distinguish between longer paraphrases that provide context and rationale for a question, and shorter, de-contextualized paraphrases suitable for re-asking the question, ensuring the user isn’t confused by repeated information when the question is re-asked (Image 1.8).

Image 1.9 shows how question paraphrases are used in a chat.

Best Practice Design Tips to Ensure Emotionality

In human-human interactions, emojis convey subtle emotions and add depth to conversations. Similarly, AI can use emojis to express emotional nuances, making interactions feel more personal and relatable. Additionally, AI should interpret emojis in user input to recognize emotional cues and respond accordingly.

In Juji Studio, designers can add emojis to a chatbot’s message by copying and pasting the corresponding Unicode emoji character, e.g., (Image 2.1).

Juji AI can automatically interpret emojis in user input and generate appropriate responses (Image 2.2).

To recognize a specific user emoji and respond with a custom message, create a "trigger" under the “Customize Chatbot Response” section to enable keyword matching with a specific emoji and define a custom response when such an emoji appears (Image 2.3).

Visual elements like images and videos make AI-human conversations more engaging and effective. They provide context, make explanations clearer, and add dynamism to interactions, keeping users engaged and improving their overall experience.

In Juji Studio, designers can incorporate visual elements in two ways: inserting an inline visual element or an independent visual message card. Specifically, designers can insert an image as part of a chatbot message (Image 2.4) or create a “Web Media” card to display images or videos individually or in a carousel (Image 2.5).

Empathy builds deeper connections in human interactions, making individuals feel understood and valued. Similarly, when AI recognizes and responds to user sentiment empathetically, it creates emotional connections, leading to more productive and positive exchanges.

In Juji Studio, designers can easily configure a chatbot to detect and acknowledge user sentiment in a controlled, custom way (Image 2.6), demonstrating AI empathy.

Best Practice Design Tips to Ensure Transparency

One of the guidelines to ensure transparency is that a chatbot should be upfront about its purpose and capabilities to set clear expectations and build trust with its users.

Juji Studio supports designers to do so in two ways. First, when a designer instructs Juji to generate an AI chatbot, they will be asked to explicitly state the following information: the organization that the chatbot is created for, the target audience of the chatbot, and the purpose of the chatbot (Image 3.1).

Given this information, Juji automatically generates an AI chatbot including its "Welcome" message (Image 3.2). As displayed in Image 3.2, the generated "Welcome" message, which the chatbot uses to start a conversation with a user, clearly explains the chatbot’s purpose and capabilities, setting the user's expectations.

Second, experienced designers can either craft their own welcome messages or customize the AI-generated message (Image 3.3) to further enforce the AI design best practice of being transparent about the AI's purpose and capabilities.

Another guideline to ensure transparency is to keep users informed on what the chatbot can do. This is especially important for proactive chatbots, which are designed to actively guide users to achieve their goals. Here we discuss three scenarios:

Suggest Starter Topics to Jump Start User-AI Chat
When a user meets with a chatbot for the first time, the user might not know what the chatbot can do. It's a good practice for designers to teach the chatbot to suggest a hand-picked set of initial/popular topics (Image 3.4). Even better, the starter topics can be dynamically selected based on certain criteria, e.g., suggesting the most popular topics or the newest topics that the chatbot just learned (Image 3.5). Just like the road signs on a map, these starter topics enable users to start their exploration.

Suggest Relevant Topics to Enable Users to Learn More
When a user asks a question and a chatbot answers, it is often the case that the user may want to know more about relevant topics. It would be helpful for a proactive AI chatbot to present a few relevant topics after answering the user's question. This enables the user to learn more (Image 3.6). In the case where a user's inquiry is vague and ambiguous, a proactive AI chatbot can suggest relevant topics for the user to clarify their intention (Image 3.7). This approach beats asking a user to rephrase their requests BLINDLY without guidance.

Suggest Related Topics to Enable Users to Learn More Deeply
Sometimes, a topic may be complex, involving several sub-topics. In such a case, a proactive AI chatbot could better guide a user to drill down on the specific information more effectively by eliciting additional user input to better understand their needs and then recommend related topics accordingly (Image 3.8).

All the best practices mentioned above are already built into Juji Studio and configurable by designers, who can then instill more transparency into their AI chatbots.

In summary, by ensuring naturalness, emotionality, and transparency in your chatbot design, you can create more engaging user experiences, making AI-human interactions more effective and enjoyable. Juji Studio simplifies the design process even for novice conversational AI designers, fostering user trust and satisfaction from the very start of an AI-human engagement.

How to Select a Chatbot Platform Powered by Generative AI

2024-06-05T17:37:00Z

Are you considering an AI chatbot for your business? A simple Google search will present you with hundreds if not thousands of AI chatbot platforms! Almost all of them claim that they are powered by generative AI and all seem to offer similar features, giving the first impression that any of them would do the job.

In reality, not every chatbot platform is created equal or is a perfect fit for the tasks you have at hand. For example, many of our customers mentioned to us before they found Juji, they had explored 20+ chatbot platforms but failed to find the one that meets their needs.

To help you expedite your chatbot platform selection process, in this article we will first identify different types of chatbots based on business needs, and then dive into each type of chatbot, listing the relevant key technical and business factors you should consider when evaluating different AI chatbot platforms for supporting that specific type of chatbot.

Identify the type of chatbot based on your business needs

First of all, business needs dictate which type of chatbot is needed.

To select a chatbot platform, the very first step is to understand why you need a chatbot and the expected tasks to be performed by the chatbot. You can then identify the type of chatbot that is needed. Based on the main tasks to be accomplished, chatbots can be roughly categorized into three types:

1. Question-Answering Chatbots (QA Chatbots)

The main task for these chatbots is to answer user questions based on public and/or proprietary information. For example, universities may use a chatbot on their website to answer student questions about learning programs (see Image 1.1 below), or hospitals may use a chatbot in their mobile app to answer patient questions on medical services. Very similar to the behavior of ChatGPT, this type of chatbot is typically driven by user questions and assumes that all questions can be answered in a SINGLE response, which is not the case in many real-world applications.

2. Question-Asking Chatbots (Survey/Interview Chatbots)

The main task for these chatbots is to ask user questions and gather information from an intended audience. Very similar to an online survey, this type of chatbot has been used to gather information for research studies, personality assessments, and job interviews (Image 1.2 provides an example of an interview chatbot). Typically, this type of chatbot is NOT capable of answering user questions including clarification questions during the process.

3. Proactive, Two-Way Conversational Chatbots (Proactive Chatbots)

In real-world applications, chatbots often need to perform multiple tasks simultaneously to deliver desired user experience and accomplish the business objectives. For example, a QA-only chatbot on a university website may be inadequate because the same chatbot must also ask questions to gather necessary information from the visitors before it can answer their questions (see Image 1.3 below). Similarly, a survey-only chatbot for job interviews may be unsatisfactory because it must also answer candidate questions regarding the positions and the organization during the interview. This third type of chatbot is a hybrid of the QA and survey chatbots, capable of both asking and answering questions during a conversation, known as Proactive Chatbots.

Building QA Chatbots

Here we outline 3 key factors to consider when selecting an AI chatbot platform to build a Question-Answering Chatbot (QA Chatbot).

Answer quality and quality control

A QA chatbot's main job is to answer user questions. When evaluating an AI chatbot platform, the first factor is to evaluate how well the platform enables a chatbot to answer user questions accurately with no or minimum AI hallucinations. Unfortunately, the direct use of generative AI technologies, like Large Language Models (LLMs) and Retrieval Augmented Generation (RAG), can't always guarantee high-quality AI responses because of their probabilistic nature. Therefore, ask the following questions during evaluation:

How does your AI platform ensure the quality of AI responses, e.g., avoiding AI hallucinations?
What kind of control measures can humans use to fine-tune AI responses (see examples in Image 2.1 below)?

Handling complex Q&A

Most AI chatbots powered by generative AI can answer simple questions—questions that can be answered with a single response. However, if your business requires a chatbot to handle complex user questions, it is important to evaluate how well an AI chatbot platform can support complex Q&A. In particular, ask the following questions during the evaluation:

Can your AI platform handle complex Q&A, which involves multi-step interactions between a chatbot and a user (see Image 2.2 below for an example)?
Who will define and manage the complex Q&A and how? For example, during a complex Q&A, the user may trigger another complex or simple Q&A, how will it be handled?

Handling complex Q&A is a very difficult problem. So you want an AI platform that is capable of handling various complex Q&A flows automatically.

Real-time chatbot Q&A monitoring and improvements

AI is far from perfect. When evaluating an AI chatbot platform, you want built-in tools that can monitor your chatbot behavior in real time, alert you with missed or problematic Q&As, and you can improve the chatbot in real time without interrupting any ongoing conversations (read more on this aspect at https://juji.io/blog/q-a-dashboard/

Building Survey/Interview Chatbots

Building an effective chatbot to ask questions and conduct surveys or interviews requires much more than just compiling a list of questions. Here are 3 critical factors to consider when selecting an AI platform to build a survey/interview chatbot:

Abilities to elicit quality user responses

The main task of a survey/interview chatbot is to gather quality responses from target users. When evaluating an AI platform, it is important to look for the following abilities:

Handling users' irrelevant, nonsense responses. An effective chatbot should recognize user responses that are off-topic or gibberish (Image 3.1), and prompt the user for a more relevant answer.

Managing users' vague or brief responses with follow-up questions. To elicit quality responses, an interview chatbot should do what a human interviewer would: ask follow-up questions to drill-down to details if user responses are vague or brief (Images 3.2-3.3 show how to instruct follow-up question generation and examples of the follow-up questions generated).

Question configuration: Not all survey/interview questions are equal: some are to establish rapport, some are required, and some are not. Just like a good survey platform, a good AI platform should allow a chatbot designer to configure each chatbot question, e.g., whether a question is required (Image 3.4).

Abilities to handle user digressions

Each person has their own unique way of thinking and conversing. Even when there's a set agenda, human-to-human conversations rarely follow it precisely. The same goes for AI-human interviews. Therefore, your AI chatbot should be able to handle diverse digression situations, to bring the conversation back on track and successfully complete a survey/interview.

Since handling diverse types of digressions requires considerable effort, look for an AI platform that offers automatic, configurable digression handling (Image 3.5), accelerating time-to-value and AI ROI.

Abilities to generate survey/interview summary reports and analytics

It's time consuming to wade through hundreds or thousands of chat transcripts and distill interview insights. When looking for an AI platform to build interview/survey chatbots, you want a platform that can also generate interview insights for fast and easier consumption, e.g., graphs and charts to illustrate trends and patterns of quantitative responses and summary of free-text responses (Image 3.6).

Additionally, an AI platform is much more desirable if it can automatically infer and analyze characteristics of survey/interview participants, which can deepen interview insights, such as reliability of the responses (see Images 3.7-3.8 for inferred characteristics of interview/survey participants).

Building Proactive Chatbots

Unlike QA-only and survey-only chatbots, proactive chatbots support two-way AI-human conversations, both asking and answering user questions during a chat. They typically are used to automate complex task flows that handle the interleaving of multiple, smaller tasks, such as answering a user's questions in the middle of interviewing the user. These chatbots are essential for automating and scaling services that involve high-touch, complex personal interactions (see the blog on this topic at https://juji.io/blog/do-you-really-need-an-ai-chatbot-use-these-3-questions-to-decide/). Since proactive chatbots are a hybrid of QA chatbots and survey/interview chatbots, all the factors discussed earlier in this article still apply. However, proactive chatbots are more than a simple combination of two types of chatbots, 4 additional factors should be considered when selecting an AI chatbot platform to develop a proactive chatbot.

Interactional intelligence

In addition to language skills powered by generative AI (LLMs), a proactive chatbot must also have interactional intelligence that enables it to concentrate on the task at hand while making users feel heard and delivering a pleasant and effective engagement. It actively engages with users, guiding and nudging them to accomplish complex tasks (e.g., tutoring a user to learn a skill or advising a user to come up with an individualized career development plan). It handles user interruptions gracefully and diligently maintains a conversation agenda to ensure task completion (Image 4.1). Most platforms require chatbot designers to manually track and maintain conversation flow in order to handle disruptions or exceptions, which is impractical given the number of potential interruptions that may occur anytime at any depth of a conversation. Thus, it's crucial that a chatbot platform is equipped with built-in interactional intelligence.

Personal intelligence

Since a proactive chatbot is meant to automate high-touch services, it is also required to have personal intelligence: the abilities to read between the lines to infer the users' unspoken needs and wants from a conversation, and personalize each interaction accordingly to better guide/persuade the users to make complex decisions or take proper actions. For example, an AI care assistant can infer a patient's motivations and personality and persuade them to adhere to care routines with stories that they can easily relate to and be motivated by (Image 4.2). It is non-trivial to power AI chatbots with personal intelligence, so make sure to ask whether and how your chatbot can acquire personal intelligence when selecting an AI chatbot platform.

Prudence

Since proactive AI chatbots leverage generative AI, unavoidably, they face the same AI safety challenges faced by any other generative AI application. Because of their complex nature and often sensitive use cases (e.g., aiding human essential services in education and healthcare), it is paramount that organizations looking to adopt proactive AI chatbots must have controls and governance tools in place for the safe use of such chatbots. When looking for AI chatbot platforms to build proactive AI chatbots, make sure to ask whether and what kind of AI safety control measures and configurations are in place to ensure safe applications of proactive AI chatbots (Image 4.3).

Plug-n-play

Building a proactive AI chatbot with interactional intelligence, personal intelligence, and prudence from scratch may require deep AI + IT expertise and cost millions of dollars and multiple years of efforts. When selecting an AI platform for building proactive AI chatbots, make sure to ask who would need to build, customize, operate, and maintain such a chatbot and typically how long it takes to do so. Built-in AI chatbot templates/assets and fast, no-code customization are always desired (Image 4.4), which in turn significantly impact time-to-value and AI ROI.

Conclusion

By identifying the right type of chatbots for your business needs and considering the relevant key factors outlined in this article, you should be able to select an AI chatbot platform that enables the development of effective chatbots to enhance your service offerings.

Do You Really Need an AI Chatbot? Use These 3 Questions to Decide

2024-06-04T18:10:01Z

With generative AI becoming a hot topic (again), CXOs started adding AI projects, especially implementing an AI chatbot like ChatGPT for business, on their to-do list. While a chatbot powered by generative AI may be a game-changer for business, investing in a chatbot is not cheap nor trivial. So it is worth asking first: does your business really need an AI chatbot? We suggest that business leaders answer the 3 questions below to quickly determine whether it is worth investing in a chatbot for a line of business.

Q1: Do you want to improve profit margins of services that involve labor-intensive, time-consuming human interactions?

If your answer is no, you don't need a chatbot.
If your answer is yes, it's worth considering.

Chatbots are meant to automate human interactions, especially those that are labor intensive and time consuming. Using the enrollment counseling service as an example, an enrollment ounselor must handle complex, personal interactions, such as guiding prospective students to explore suitable program options, addressing their concerns, and eventually helping them make an enrollment decision. If your service is like counseling and tutoring services that require much human time and effort, investing in a chatbot might be worthwhile: adopt a chatbot to automate human interactions and reduce human workload.

Q2: Do you want to scale out services that require extensive human effort and time?

If your answer is no, you don't need a chatbot.
If your answer is yes, it's worth considering.

High-touch servies like advising and counseling are hard to scale due to the human and time effort required. Hiring more staff may not even be a solution. For example, prospective students of online programs who have full-time jobs or are located across the globe, need assistance beyond regular headquarters' business hours. Hiring human staff to cover off-hour time frames can be extremely costly and impractical. On the other hand, an AI chatbot can operate 24/7 and interact with hundreds of thousands of users at the same time.

Q3: Do you want to have a happy and healthy workforce?

If your answer is no, you don't need a chatbot.
If your answer is yes, it's worth considering.

Humans dislike repetitive tasks. For example, a human tutor is not thrilled to answer the same question or deliver the same tutorial for hundreds of students over and over again. Emotionally taxing tasks can also lead to burnout, as seen with grievance counselors. In contrast, an AI chatbot can take on the repetitive work without a complaint or emotionally demanding tasks without being burnt out. Humans can then focus on more meaningful or pleasant tasks, thereby enhancing job satisfaction.

In summary, if your answers are "yes" to all above 3 questions, it is worth considering the adoption of an AI chatbot. If you want your chatbot to get its job done, you want a chatbot powered by generative AI as a minimal requirement.

Put It in Writing: 3 Reasons Why ChatGPT Plus Inadequate for Creating a Business Chatbot

2024-05-28T22:15:22Z

With the launch of ChatGPT plus, more and more organizations started to build their own chatbots. While ChatGPT Plus provides a natural language interface for almost anyone to build a chatbot, it is inadequate for building a chatbot in a business context.

Why? The rationale is quite straightforward. In a business context, when we negotiate a business contract or agreement, we often say "put it in writing" for the purpose of recording important terms and make them official. The same principle should also apply to regulating an AI chatbot's behavior especially when the chatbot is to act on behalf of an organization or business.

Unfortunately, a prompt-driven tool like ChatGPT Plus does not put an AI chatbot’s behavior in writing, hence it is inadequate for building a chatbot for business use.

In the rest of the blog, we outline 3 reasons why it is important to put AI chatbot behavior in writing, especially in a business context.

Reason 1: Ensure the integrity of AI chatbot behavior

AI is far from perfect and makes mistakes from time to time. Let's assume that a health science researcher uses ChatGPT Plus to create an AI chatbot to interview patients and gather critical patient information. If such an AI chatbot is created and deployed solely based on the researcher's prompt describing the interview process, the AI chatbot will most likely behave differently with each interviewee since each conversation will be generated on the fly, and there is NO guarantee that the AI chatbot will even ask the same question in every conversation!

Let's look at another example in education where an AI chatbot is created to give interactive tutorials on certain learning content. If an instructor just uses a prompt to instruct an AI chatbot to deliver such a tutorial, the AI chatbot may not even follow the same tutorial steps with each student.

The culprit of such inconsistent AI behavior is the nature of large language models (LLMs) like GPT because they produce behavior (e.g., questions to ask users or answers to give users) based on probability distributions, which imply inherent uncertainty.

If an AI chatbot is created for entertainment purpose, inconsistent or uncertain AI behavior may be tolerated or even be a good thing. However it is unacceptable for conducting a rigorous scientific research study or conducting organization communications, where AI behavior should be consistent and appropriate at all times.

To ensure the integrity or consistency of AI chatbot or chatbot behavior, a good practice is to put the AI chatbot's behavior often in the form of a workflow in writing to make sure it follows the flow and behaves consistently with all users.

Fig 1 below shows an example prompt used to generate a chatbot's workflow automatically.

Fig 2 records the generated workflow, which allows a human designer to review and revise as needed. With the workflow in writing, the chatbot will follow the workflow consistently to interact with each user.

Fig 3 shows another example prompt that is used to generate a multi-step, interactive tutorial based on given learning content, automatically.

Fig 4 shows the generated tutorial steps in writing. Again, the recorded steps allow a human designer to review, revise, and approve the generated tutorials.

Reason 2: Ensure Deep Controls of AI Chatbot Behavior

Above we mentioned that documenting an AI chatbot’s behavior should document its workflow. The workflow however is a static representation of a chatbot’s behavior and there is AI logic behind the scene to drive the chatbot to execute its workflow when it interacts with a user. In other words, there is much more than meets the eye when designing a chatbot’s workflow, especially in a business context.

Let's use the same example of creating an AI chatbot to serve as a patient care assistant, which will check-in on each patient daily and gather critical patient information. While tools like ChatGPT can easily generate a set of questions for such an AI chatbot to ask a patient and gather patient information, it is unclear what kind of behavioral logic drives the chatbot to conduct the check-in properly and successfully.

For example, does the existing AI behavioral logic handle the following situations:

How does the chatbot handle the situation if a user is unwilling to answer the question?
How does the chatbot handle the situation if a user does not know the answer?
How does the chatbot handle the situation if a user gives only very terse input?
How does the chatbot handle the situation if a user expresses depressions or is in a bad mood?
How does the chatbot handle the situation if a user asks for medical advice?

We can go on and on with such "if" statements. Just like that every organization has their logic of doing business or handling communications, which often are spelled out and written down, organizations should also demand to have their AI chatbot behavioral logic spelled out and written down. More importantly, organizations should customize the AI logic to suit their business needs or to fulfill their compliance requirements.

Instead of requiring an organization to come up with the difficult-to-anticipate AI behavioral logic from scratch, Juji has distilled the most important set of AI behavioral logic, established optimal key configurations for a number of AI chatbot situations, and spelled out such a configuration for an organization to review and revise as needed.

In other words, when you generate an AI chatbot in Juji Studio, you get the chatbot workflow (see our yesterday's post) with pre-built AI behavior logic, all spelled out and customizable. Truly more than meets the eye.

Juji supports a human designer to control an AI chatbot's behavior at 3 levels below.

Fig 5 below shows various chatbot controls that a human designer can use to control a chatbot's overall behavior and logic.

Fig 6 shows various chatbot controls that human designers can configure to control the chatbot behavior and logic inside a particular conversation topic.

Fig. 7 below shows controls that human designers can use to control a chatbot's behavior based on a specific type of user input.

In short, "put it in writing" also means spelling out the behavioral logic of an AI chatbots so one can view and deeply control how AI behaves if needed.

Reason 3: Ensure the business provenance of an AI chatbot (AI chatbot Provenance)

Business provenance refers to the documentation and verification of the origin, history, and custody of business-related data, processes, and products. It is often required for various reasons (e.g., accountability and compliance). When an AI chatbot represents a business to engage with its customers, employees, and partners alike, this AI chatbot along with its associated assets, such as the avatar and the knowledge, all become business-related data, processes, and products, hence subjecting to business provenance.

We use the term "AI chatbot Provenance" to refer to the process of documenting, verifying, and monitoring the origin, history, and custody of AI chatbots and assets.

To facilitate AI chatbot provenance, Juji has provided an array of tools in the following 4 areas:

(a) Trust and Transparency

In Juji, an AI chatbot's design, such as its general settings, workflows and knowledge used to answer user questions are explicitly written out and easily accessible and configurable by non-IT subject matter experts (SME).

Fig 8 shows various configurations that non-IT SMEs can access and control.

Moreover, such business data and processes can be easily shared within an enterprise for further scrutiny and validation. Fig 9 below shows that one can easily share an AI chatbot along with its assets with colleagues for added trust and transparency during the design and operation of the chatbot.

(b) Quality Control

No AI chatbots are perfect. While a collaborative AI chatbot design/development process helps improve the quality of AI chatbot, continuously monitoring AI chatbot's behavior is another critical step to AI chatbot provenance. Juji helps businesses monitor AI chatbot behavior continuously and creates a to-do list to alert human custody to audit and improve their AI chatbots accordingly (see Fig 10 below).

Just like products and processes in established industries such as pharmaceutical and Finance, AI chatbots as a new, imperfect product will surely be subject to regulatory compliance such as President Biden's executive order on generative AI. Juji's collaborative AI chatbot platform facilitates regulatory compliance as it enables non-IT professionals such as legal professionals to be part of the AI chatbot design and development process. Juji's multi-level AI safety measures also aid regulatory compliance.

For example, Fig 11 below shows examples of AI responses labeled per their content origin.

(d) Risk Management

Juji's AI chatbot behavioral logs (e.g., workflows, chat transcripts, chat summaries, and to-do list) aid crisis management (e.g., detecting AI chatbot misbehavior timely) and prevent fraud (e.g., checking the history of saved or exported AI chatbot dossier to trace the history of an AI chatbot).

5 Design Tips for Conversational AI Designers to Make a Proactive AI Agent

2024-04-27T01:54:34Z

Recently, we have heard more people, especially top technology experts, talk about creating "proactive" AI agents with agentic workflows (e.g. see a recent video by Andrew Ng). Designing a proactive AI agent requires skills. In this article, we present 5 design tips for conversational AI designers to make effective proactive AI agents. In each design tip, we use concrete examples to explain the desired AI agent behavior and show the corresponding Juji features you can use to enable such a behavior.

Tip 1: Make an articulate AI agent

A proactive agent doesn’t just passively respond to user questions. Rather, it actively engages with a user and guides the user toward an agenda (e.g., tutoring the user to learn a skill or coaching the user on coming up with an individualized career development plan). During such an engagement, just like a good human agent, a proactive agent should be articulate in expressing itself, no matter whether it asks or answers a question. Below are two situations where a proactive agent should be articulate in making a conversation more natural and fluent.

(1) In many applications including interviewing and counseling, a proactive AI agent often needs to ask user questions to elicit necessary information before guiding the user toward a goal. Just like a human interviewer, an effective AI agent should be able to ask the SAME question in different ways depending on the context. The image below (Image 1.1) shows a conversation where an AI agent must repeat its question when the user didn't answer the question.

Humans will never re-ask the question using exactly the same expression. Neither should the AI agent. The next image (Image 1.2) shows that the AI agent re-asks the same question in an alternative expression, which makes the conversation much more natural.

A conversational AI designer can use Juji Studio to customize an AI agent’s message, including paraphrasing a question either manually or automatically using generative AI (Image 1.3).

Furthermore, for each expression, a conversational AI designer can indicate whether it is only used for re-asking, giving you fine-grained control to make an AI agent articulate (Image 1.4).

(2) In a situation where a proactive AI agent must ask a series of questions to elicit information, the AI agent should also be able to ask these questions in a conversational tone and with smooth transitions from one question to the next, similar to what an articulate human interviewer would do.

In Juji Studio, conversational AI designers can use generative AI to auto-generate a conversation flow with multiple questions (Image 1.5), which you may edit with more human touch as needed.

The result is a fluent conversation flow that resembles those of human-human conversations (Image 1.6).

Tip 2: Make a well-prepared AI agent

Human conversations are highly diverse and rarely single threaded, which require active listening, empathy, and negotiation skills to keep a conversation going and achieve the intended communication goals. Just like a good human conversationalist, a proactive AI agent should also be well-prepared to handle diverse conversation situations gracefully. Below we list several frequent conversation situations where a proactive agent should be well-prepared to handle them.

(1) Handle user reciprocal questions. In a human-human conversation, humans often ask reciprocal questions such as "What's yours" and "what about you". A well-prepared AI agent should also handle user reciprocal questions (Image 2.1).

(2) Handle user clarification questions. Different people have different background and knowledge. A perfectly clear question to one person may be difficult for another to understand. So prepare your AI agent to answer user clarification questions during a conversation (Image 2.2).

(3) Handle user excuses. Just like in human-human conversations, people may dodge a question by giving various excuses. A proactive AI agent should also be prepared to handle diverse user excuses to accomplish an agenda, e.g., eliciting quality information from a target audience (Image 2.3).

(4) Handle related user questions in context. During human-human conversations, people may ask relevant questions to a topic under discussion. A proactive AI agent should also anticipate relevant user questions and be prepared to answer the questions in a conversation (Image 2.4).

Feel overwhelmed that you have to handle all the above situations and don't know how? No worries, conversational AI designers can use Juji Studio to prepare an AI agent to handle all the situations mentioned effortlessly with 100% #NOCODE. For example, A designer can easily toggle options to decide whether to accept a user's "I don't know" excuse or gibberish input in a specific conversation topic (Image 2.5).

A designer can also upload one or more proprietary data and then ask Juji to generate a set of Q&As automatically to handle user questions related to a topic (Image 2.6).

Tip 3: Make a responsible AI agent

Every AI agent is designed with a purpose. A proactive AI agent must act responsibly to ensure task completion and goal fulfillment. So how can conversational AI designers instill a sense of responsibility into an AI agent? From our own experiences, we have found there are at least three types of responsibility that a proactive AI Agent should have.

(1) Ask follow-up questions to deepen a conversation and accomplish the goal. In many applications such as interviewing and counseling, a proactive AI agent typically asks open-ended questions to elicit richer information. However, users may not cooperate and instead provide overly general or vague information. In such a situation, a responsible AI agent should ask follow-up questions to deepen the conversation, ensuring that relevant details are gathered. The two images below (Image 3.1 and Image 3.2) show an example where the AI agent automatically generates follow-up questions to elicit must-have and nice-to-have information based on the user's responses.

(2) Stick to an agenda for task completion. Every proactive AI agent should have an agenda (aka workflow) that drives the conversation toward task completion. However, most conversations often digress. A responsible AI agent should manage the digressions/distractions and stick to the agenda. The following image (Image 3.3) provides an example of a conversation where the AI agent handles a user digression gracefully and then goes back to the agenda. Conversational AI designers don't need to do anything but directly leverage Juji Studio's built-in capabilities for AI agents to stick to their agenda automatically.

(3) Be a good human collaborator. AI is far from perfect. There will always be questions that an AI agent can’t answer or tasks it can’t complete. A responsible AI agent should collaborate with human teammates when necessary. For example, the image below (Image 3.4) shows the interface in Juji Studio where an AI agent flags unanswered questions or unfinished tasks in a to-do list for humans to review and to improve the AI agent's performance over time.

Tip 4: Make a perceptive AI agent

Effective human communicators have great social-emotional intelligence in addition to their language skills. In other words, they can read people during a conversation and use the reading to empathize with their conversation partners and foster effective and pleasant conversations. Similarly, a proactive AI agent should be perceptive and possess the ability to actively listen to its users, infer the users' unspoken needs and wants from a conversation, and personalize each interaction. Such capability enables AI agents to better help their users based on each user's unique needs and wants. Here are two conversational design tips to power a perceptive AI agent.

(1) Enable an AI agent to read user emotions. A perceptive AI agent should actively listen to its users, echo their sentiments and recognize their feelings. This makes users feel heard and understood. The following image (Image 4.1) gives an example of an AI agent detecting sentiment from the user’s input and responding empathetically.

The image below (Image 4.2) shows how a conversational AI designer can easily customize the AI agent’s response in Juji Studio based on specific user sentiment detected during the conversation.

(2) Enable an AI agent to read user personality. A perceptive AI agent should also read between the lines in a user engagement to deeply understand each user’s unique characteristics, including his/her motivations, interests, needs, and personality. The image beneath (Image 4.3) shows how an AI agent is able to infer a user's Big 5 personality traits from their conversation. Conversational AI designers can leverage Juji Studio to detect AI-inferred user personality and generate messages based on user personality.

Tip 5: Make an adaptive AI agent

A skilled human communicator adapts each conversation to their conversation partners, making the engagements more effective. Likewise, a proactive AI agent should adapt its engagement to each of its users based on the user's background, interests, and personality. Below are two common types of adaptation that should be enabled to make an adaptive AI agent.

(1) Tailor AI agent explanations to suit an individual’s background, interests, and level of expertise to make the content more relatable and easier to understand. One of the important tasks for a proactive AI agent is to tutor students on complex concepts. If an AI agent can provide detailed explanations using terminology, analogies, and examples that each student is familiar with, it facilitates learning. The image below (Image 5.1) shows an example of an AI tutor that dynamically adapts blockchain tutorial content for a student who loves to cook.

Conversational AI designers can leverage Juji Studio to auto-generate an interactive tutorial from proprietary data (Image 5.2)

The designers can also review and revise the steps as needed (Image 5.3). During runtime, Juji’s AI agent dynamically tailors the content in each step of the tutorial to suit the user’s background and interests.

(2) Adapt suggestions/persuasions/guidance based on a person’s personality traits. Everyone is unique. If an AI agent can provide suggestions or guidance that can better resonate with each user based on their unique motivations, needs, and interests, it is more likely the user will follow the suggestions or guidance. The following image (Image 5.4) shows an adaptive AI career assistant that suggests a suitable job opening along with the required skills based on the user's goal, personality, and available job roles.

The next image (Image 5.5) shows an adaptive AI care assistant that motivates and persuades a patient to adhere to care routines with stories generated based on the patient's personality so the patient can easily relate to and be motivated by.

With Juji Studio, conversional AI designers can easily enable an AI agent to generate personalized messages based on Juji inferred user personality insights (Image 5.6).

In summary, we present 5 design tips for conversational AI designers to make proactive AI agents articulate, well-prepared, responsible, perceptive, and adaptive. Juji Studio makes it easy to create such an AI agent. Sign up to create your own and put those tips to use!

Juji Powers the University of Illinois No-Code AI Chatbot to Help Grow Student Recruitment

2024-01-22T13:55:09Z

Juji an Artificial Intelligence (AI) company that specializes in powering cognitive AI assistants in the form of chatbots,released a case study, highlighting how a Juji AI chatbot helped the University of Illinois online MBA program (iMBA) to scale out student recruitment.

“A chatbot solution could help us automate certain operations so we can meet increasing student demands while keeping our staff happy.”

“We have a handful of recruitment staff members who need to answer tens of thousands of questions from prospective students in addition to their other responsibilities such as advising students,” said Brooke Elliott, the Associate Dean of the College. “A chatbot solution could help us automate certain operations so we can meet increasing student demands while keeping our staff happy."

The Assistant Director of Admissions and Recruitment, Myranda Crist, led the team in designing and implementing an AI chatbot, called Alma, which sits on the iMBA’s main web page and performs three tasks:

Answer prospective students' free-text questions about the program
Elicit prospects' information for follow up
Inform prospects the key information about the program

The team used Juji Studio, a no-code AI chatbot builder, to customize the main chat flow of Alma based on one of Juji AI Assistant templates–the AI Program Assistant template with a pre-built chat flow and a set of program-specific FAQs to complete all three tasks. Alma was created and deployed in a couple of weeks with no IT resources involved. After Alma went live, Juji automatically captured unanswered user questions, enabling the team to rapidly update Alma without interrupting Alma's ongoing conversations with prospective students.

The study showed that among the prospective students who interacted with Alma,11.9% applied for the program, while only 6.9% of prospects who didn't chat with Alma applied for the program. Combining Myranda and team's teaching and Alma's continuous learning capabilities, Alma has also answered over 99.5% of student questions.

“Universities and their recruiting staff are overwhelmed by the volume of student inquiries, stretching resources thin,” said Michelle Zhou, co-founder and CEO of Juji. “Since human-driven conversations are hard to scale, organizations like University of Illinois are enlisting the help of no-code conversational AI to rapidly set up, deploy and manage chatbots that can streamline operations and scale out recruitment efforts to keep both prospective students and recruitment staff happy.”

"We love Alma,” noted Myranda. “It’s created another way for prospects to connect with us and receive quick answers. Alma has helped inform us of areas in our process that may not be as clear as we would like, and we’ve been able to address that. Another added benefit is that our current students also utilize Alma for answers, something we did not expect!"

Juji Launches New Generative AI Tool for Automating Cognitive AI Chatbot Creation

2024-01-16T09:14:43Z

Juji announced the launch of its new generative AI tool that enables any organization to rapidly generate custom AI chatbots with no code or additional IT resources. These AI chatbots are powered with cognitive intelligence—advanced human soft skills like active listening and reading between the lines, and can be used to scale human teams and automate high-touch, high-stakes engagements 24x7.

Unlike other generative AI platforms, which create static content like text and images, Juji can generate autonomous AI chatbots that are customized to a specific workflow and proprietary knowledge

"Unlike other generative AI platforms, which create content like text and images, Juji can generate autonomous AI chatbots that are customized to a specific workflow and proprietary knowledge," said Dr. Huahai Yang, co-founder and CTO of Juji. “While other chatbot companies focus on using generative AI technology like GPT-4 to teach chatbots language skills, Juji uniquely combines generative AI with computational psychology to power chatbots with advanced human soft skills in addition to language skills. As a result, Juji chatbots behave empathetically and responsibly."

One-on-one, high-touch engagements help organizations aid their prospects and customers in high-stakes decision making, but they are difficult to scale. With Juji's generative AI tool, organizations across industries including education, talent management and healthcare can now auto-generate custom AI chatbots rapidly to scale out such engagements. These generated Juji AI chatbots can proactively engage with their users in a two-way conversation, infer their personality traits in real-time, and guide them to make important life decisions based on their unique personality (e.g., choosing a career).

The global generative AI market size is estimated at $8.65 billion in 2022, growing to $168.62 billion by 2032. While organizations are eager to harness the power of generative AI like ChatGPT, they are also concerned with the how-tos and risks.

“Generative AI is very powerful. However, not every organization knows how to leverage the power.” said Dr. Michelle Zhou, co-founder and CEO of Juji, “Not only does our new tool enable any organizations, with or without AI or IT expertise, to harness the power of generative AI in a unique way and safely, but it also enables organizations to automate high-value engagements with empathy. This is a significant advancement towards democratizing AI and bridging the AI divide."

Juji Powers Five-University Study on Validating Personality Scores Inferred by an AI Chatbot

2024-01-07T13:45:37Z

Juji, Inc., an Artificial Intelligence (AI) company that enables auto-generation and operation of cognitive AI chatbots, announced the findings of a study, “How Well Can an AI Chatbot Infer Personality? Examining Psychometric Properties of Machine-inferred Personality Scores.” The study examined the quality of AI-inferred personality scores. Instead of asking participants to take a traditional personality test, in this study, an AI chatbot built on the Juji platform engaged with 1,440 people in a one-on-one, two-way conversation and then derived each person's personality scores from the conversation in real time.

“I was initially skeptical about AI-inferred personality traits and their effectiveness. Now I'm excited about the findings and the process of how AI works”

The study presents the most comprehensive validation of AI-inferred personality scores to date. "The study did an excellent job examining the psychometric properties of the AI-inferred personality scores—very comprehensive in scope," commented Professor Lillian Eby, editor-in-chief of the Journal of Applied Psychology where the study has been accepted to appear.

“I was initially skeptical about AI-inferred personality traits and their effectiveness. Now I'm excited about the findings and the process of how AI works,” said Dr. Jinyan Fan, the lead researcher and a professor in the Department of Psychological Sciences at Auburn University, specializing in talent selection and retention. “The use of no-code AI chatbots to derive personality from conversations presents a new, effective, and more natural and empathetic form of psychometric assessment. It can help scale out human-driven interviews and personality insights discovery.”

This study showcases the practical value of AI-inferred personality for a wide range of applications, such as assessing student-program fit and individualizing student retention in education, workplace talent management, and aiding patient-caregiver matching.

“We are thrilled to enable such an important study, which has far-reaching implications for the entire field of computational psychology as it sets the gold standard on how to validate AI-inferred personality scores," said Juji CEO and Co-Founder Dr. Michelle Zhou. “Coupling AI-inferred personality with generative AI like ChatGPT unlocks the holy grail of human-computer interaction—AI beings can now deeply read and engage with each user at scale.”

Why Cognitive AI is Essential in Higher Education

2023-11-14T21:34:34Z

What is "conversational"? By the Merriam-Webster dictionary, conversation is defined as an "oral exchange of sentiments, observations, opinions, or ideas". Because of the exchange, one-on-one conversations are often considered the most effective way for humans to engage with one another.

However, human-driven conversations don't scale. From our conversations with universities, we learn that it is not unusual for an admissions office to have only four to five staff members, who might need to handle over 30,000 to 40,0000 incoming student inquiries, not to mention that many such inquiries are repetitive on the similar topic , e.g., program requirements, financial aid options, and application process. Additionally, universities wish to learn more about prospective students who made such inquiries, such as their interest and expectations, which will enable university staff to better follow up with the students (e.g., inviting them to apply for a program or scheduling an interview). So what capabilities must a chatbot have in order to help universities automate such engagements but without losing a human touch?

What is a Conversational Chatbot

To scale out human-driven conversations while containing cost, more and more universities now are enlisting the help of chatbots. Now you might be puzzled: how could a chatbot be not conversational? From the literal sense, it makes absolutely no sense if a chatbot is not conversational. In reality, however most chatbots out there unfortunately cannot converse with their users, hence are not conversational.

How can chatbots be made conversational? To possess human conversation skills, conversational chatbots must be powered by cognitive intelligence—an advanced, human-like Artificial Intelligence (AI). Such a chatbot, also known as a cognitive AI assistant or a cognitive AI chatbot, should demonstrate three key conversation capabilities.

A Conversational Chatbot Supports Two-Way Communication

We’ve all had the experience with a chatbot where we are clicking buttons and being forced to follow a path that leads nowhere. Frustrated, we give up and try to ask a question directly in the chat window. The chatbot, however confused, responds that it does not understand our input. This is a typical one-way exchange, where a chatbot tries to dictate the whole conversation by following a rigid, pre-defined chat flow, which often leaves users frustrated and unhelped.

In another scenario, a chatbot might open up a conversation with a phrase like "How may I help you?" What happens next is also quite familiar: after texting a question, we receive a chatbot response "Sorry I'm not sure I understand, would you mind rephrasing your question?" Left clueless, we don't know how to rephrase the question and are also afraid no matter how we rephrase the question, the chatbot still could not understand. So we give up. This is another typical one-way exchange, where a chabot tries to let users drive the conversation without providing any guidance, which often results in failed conversations because of today's technology limitations.

In contrast, conversational chatbots support two-way exchanges, delivering more human-like conversational experiences. Specifically, conversational chatbots can take initiatives and guide a conversation while actively listening to users and allowing users to take initiatives during the conversation. The example below shows a two-way conversation between a chatbot and a student who is inquiring about financial aid options.

A Conversational Chatbot Understands and Maintains Conversation Context

Continuing our above example, a typical chatbot might not recognize and respond to the user's question, even if it did, it would get lost by the user interruption or take us back to the beginning of the conversation. Either way would derail the conversation and fail to complete the intended task (in this case, recommending financial aid programs).

Just like humans in a conversation, conversational chatbots are able to understand the focus of attention and remember the conversation context. As shown below, while a user diverts from the initial flow of the conversation, the chatbot answers the question, and then naturally resumes the flow where it was left off. This conversation capability enables a chatbot to complete assigned tasks while helping users whenever possible.

A Conversational Chatbot Can Read People and Personalize Each Engagement

In addition to language abilities, great human conversationalists have high social-emotional intelligence, which enables them to deeply understand and empathize with their conversation partners and foster effective and pleasant conversations.

Unlike a typical chatbot that often offers one-size-fits-all responses and knows little about its users, a conversational chatbot can actively listen to its users, infer the users' unspoken needs and wants from a conversation, and personalize each conversation. Not only does this capability allow a conversational chatbot to behave more human-like, but also enables it to better help its users based on the users' needs and wants. The conversation example below shows that such a chatbot provides personalized career information to students based on the students' passions, interests, and personality.

Why Adopt a Conversational Chatbot Now?

The demand for personalized, conversational interactions is a growing expectation by today's university students. Multiple university studies show that students are willing to and wish to interact with a chatbot and can also benefit from such interactions. Higher education needs to cut through the clutter and create a better connection with their audience, e.g., current and potential students and alumni. Additionally, university studies indicate that technologies already exist to support rapid, no-coding setup and adoption of cognitive AI assistants, which not only deliver a higher engagement rate, but also continually improve from users' interactions.

In other words, conversational chatbot solutions are no longer a luxury item for only certain universities. Without requiring additional technical resources, EVERY university now can easily adopt cognitive AI technology and their human-like conversational capabilities to improve student/alumni engagements, reduce administrative burdens, and elevate workplace morale, as the university hopes to achieve.

AI Hallucinations could usher in the next era of AI: cognitive AI

2023-09-01T10:30:00Z

Humanity’s shiny new toy, generative AI like ChatGPT, has shown to be very powerful at interpreting users’ natural language expressions to quickly produce a wide range of text responses, from copy-editing text to summarizing a dialog. While AI responses sound quite confident, we have gradually discovered the flaws in those responses. For example, I asked ChatGPT, “Who invented IBM System U and later known as IBM Personality Insights?” It gave me the following response with names of two people, who may or may not even exist. This phenomenon is referred to as “AI hallucination.”

But are AI hallucinations all bad?

Before answering, let’s take a quick look at what causes AI hallucinations. In essence, language-based generative AI, the technology behind tools like ChatGPT, learns language patterns and structures from its training data and then generates new content with similar patterns and structures. If the training data or learning process is inadequate or flawed, inaccurate content, such as what we see above, is generated. In the above example, ChatGPT most likely does not have training data on the inventors of IBM System U or Watson Personality Insights, so it manufactured the answer by piecing together Dr. Talbot and Dr. Aric, and IBM System U based on a probability.

Although ChatGPT made a mistake in piecing together irrelevant information, its abilities of applying what it has learned from its training data to synthesize new information is somewhat related to humans’ capabilities of “learning by analogy” or “applying a solution of one problem to another situation.” These unique cognitive abilities are what set humans apart from other animals.

As machines gain more and more advanced human cognitive skills, such as drawing proper parallels and reading between the lines, we are in fact witnessing the emergence of a new AI era beyond generative AI: cognitive AI. Machines with cognitive intelligence can do wonders in many real-world applications, including education, healthcare, and career development.

Today, cognitive AI can benefit two areas directly and immediately: translating information or messages based on user characteristics to aid information comprehension and persuasion, and generating new experiences based on that of like-minded people.

FOUND IN TRANSLATION

Every human being is unique, and individual differences make people see, hear, and comprehend the world’s information differently. If AI can “translate” information in a form that a person can better relate to and comprehend, the world would be a better place.

One example is in education where many students are struggling with STEM subjects. Imagine an AI tutor that can “translate” course materials and generate quizzes—with examples or situations that a student is familiar with or can easily relate to—based on their unique psychographic characteristics, such as interests (say, sports), cognitive style (preferring visual instructions), and personality (extrovert). This personalization can then stimulate students’ learning interests and motivate interested students to learn more and better.

In healthcare, many people have difficulty digesting clinical information or following medical advice. Imagine an AI care assistant that can “translate” healthcare information or advice by using analogies and stories generated based on a person’s unique psychographic characteristics, such as their temperament (for example, impulsive and self-conscious), motivation (taking care of family), and passion (say, gardening). The following example is AI-generated, human-edited healthcare advice on obesity, tailored to a person who is a family-oriented, avid gardener.

Not only will this personalization help patients better comprehend healthcare information and guidance, but it could also motivate them to stay healthy, improving their overall care experience and outcomes.

EXPERIENCE ADAPTATION

A person’s experience is often limited due to various constraints, such as time, geographic, and financial restrictions. If AI can present potential new experiences, it could open up people’s minds as well as help them discover opportunities for optimizing their well-being.

In career development, few individuals, especially young people, have a clear idea of their future career. Imagine that an AI career counselor can understand every student’s passions, interests, talents, and skills, as well as how they handle life’s challenges. It can then help them plot and explore their unique, potential career experiences and possibilities based on the careers of other professionals who are just like them. The AI career counselor can also show students how to best capture these opportunities and maximize their potential. Below is an AI-generated, human-edited potential career experience for a young lady called Emma, who is a freshman studying computer science at the University of Texas at Austin.

Likewise, patients in a healthcare setting dealing with illness can also benefit from this technology. Imagine an AI care assistant that can deeply understand each patient, including their psychographic characteristics. It can give them insight into what their healthcare journey may look like based on the experiences of other patients who are just like them. This “experience transfer” or “experience adaptation” can give patients hope and encourage them to adhere to treatment, improving overall healthcare outcomes.

TAKEAWAYS

In order for cognitive AI to really work, we’ll need to bound the tech to a few conditions:

AI will need rich and accurate knowledge, such as the real people’s experiences.
AI will need a deep understanding of users, including psychographic characteristics.
We’ll need human experts in the loop to amend machine “incorrections” or mistakes.

We are only at the beginning of a new AI era, which has already opened the door to many possibilities that we could not have imagined just a few years ago. Hallucinations may not be a hiccup, per se; they could be the key to the next stage of AI.

Note: A version of this article is published on Fast Company

Juji Featured on AI for All: Cognitive AI, AI Assistants, and Artificial Empathy

2023-08-10T23:24:11Z

On August 10, 2023, the AI For All Podcast hosted a riveting episode featuring Dr. Michelle Zhou, Co-Founder and CEO of Juji, who joined Ryan Chacon and Neil Sahota to delve into the fascinating world of Cognitive AI.

The AI For All podcast episode delved into various aspects of cognitive AI, its benefits, and its future in enterprise.

Key discussion points included:

Cognitive AI: This technology, at its core, mimics human intelligence, helping machines understand, interpret, and respond to human behaviors.

Benefits of Cognitive AI: The technology can help organizations provide personalized services at scale, and can also perform complex tasks that are typically time-consuming for humans.

Artificial Empathy: Zhou explained the role of artificial empathy in enhancing human-machine interactions.

Robots vs Chatbots: The discussion highlighted the distinguishing features and uses of robots and chatbots, two integral components of modern AI technology.

Psychographics in AI: Psychographics, the study of people's attitudes, aspirations, and other psychological criteria, are crucial in AI to create more personalized and impactful interactions.

Adoption of Cognitive AI in Enterprises: Zhou stressed the importance of enterprises adopting cognitive AI, citing its potential to provide scalable, personalized customer service.

No-code AI: The conversation touched on the rise of no-code AI platforms, which allow individuals without coding expertise to build and customize AI applications.

AI as an Assistant: The panel discussed the role of AI as an assistant, offering help and support in a variety of contexts.

Future of AI: The discussion wrapped up with thoughts on the future of AI, with Zhou predicting a world where AI becomes an integral part of everyday life.

The full episode can be viewed online.

Can Hallucinated AI Help with Patient Engagement?

2023-08-07T11:10:00Z

ChatGPT and similar AI applications have given humankind a new tool. While this new tool is powerful, it may not always be reliable. Hence the term “AI hallucinations” is coined to refer to such unreliable AI performance.

Here is an example. I asked ChatGPT, “Who founded Juji“, the AI startup I co-founded. It hallucinated with the following reply:

It got several facts wrong including my education. I received my Ph.D. from Columbia University, not Carnegie Mellon University. Moreover, Juji was co-founded by Dr. Huahai Yang and me, not just myself.

As the medical community considers the role of generative AI, it begs the question: Can a hallucinated AI still help with high-stakes applications such as patient engagement? The short answer is yes if hallucinated AI is appropriately used.

Customize an AI Chatbot with Accurate Information

One cause behind AI hallucinations is that applications like ChatGPT are trained on public data, which may lack accurate information. For example, if ChatGPT has Juji’s proprietary data on its founding history and founder information, it probably won’t make a mistake as seen above. Similarly, reliance on trustworthy healthcare information is crucial for AI chatbots designed to engage patients. For example, an AI chatbot sitting on Mayo Clinic’s website should be trained with validated, up-to-date healthcare information to provide patients and their family members with accurate insights regarding specific diseases and treatments.

Moreover, such information must be updated from time to time to ensure its validity, especially in the healthcare industry when health situations (e.g., the COVID pandemic) or medical advances evolve rapidly.

Human-in-the-Loop Validation

Let’s assume that we have trained a healthcare AI chatbot with trustworthy healthcare information. Does this mean the AI chatbot won’t hallucinate anymore? The answer is, unfortunately, no. Occasionally, the AI chatbot may still hallucinate by piecing together irrelevant information.

Here is an example: I uploaded our company’s privacy policy into ChatGPT and asked: “Who should I contact regarding Juji’s privacy policy?” It told me to contact “[email protected],” an email address that doesn’t exist and has never been mentioned in our privacy policy either.

To further prevent or reduce AI hallucinations, it is thus critical to support a human-in-the-loop validation process. There are typically two types of human-in-the-loop validation: pre-use and post-use validations. I use a concrete example below to illustrate these two approaches.

Assume that we have provided a healthcare AI chatbot with information about various diseases like Parkinson’s and Alzheimer’s.

Based on the given disease information, a pre-use validation approach automatically generates a set of possible Q&A pairs, as shown below. A healthcare professional can then examine and validate the generated results before an AI chatbot can use them to answer user questions about various diseases.

While this approach ensures that the chatbot always provides the correct information to users, it has its limitations. For example, a user asks, “What are the symptoms of Parkinson’s disease?” and then asks, “What about Alzheimer’s?” While the chatbot can answer the first question easily based on the generated Q&A pairs (see row 1 above), it could not answer the second question because the generated Q&A would not cover such context-sensitive questions. This is because the second question uses an incomplete expression and may be interpreted differently under different contexts. The above example means, “What are the symptoms of Alzheimer’s disease?” Alternatively, if the previous question were “What are the treatment options for Parkinson’s disease?”, it would mean “What are the treatment options for Alzheimer’s disease?”

On the other hand, a post-use validation approach lets an AI chatbot directly answer user questions based on given information AND the conversation context. Using the same example above, in this approach the AI chatbot can answer both questions easily. This approach will then record user-AI interactions so a human can examine the records to confirm correct answers and rectify AI hallucinations if there are any. Although this approach may produce hallucinated AI responses at the beginning, it is quite effective at reducing and preventing hallucinations over time because it not only corrects detected hallucinations but also continuously provides verified answers to the chatbot.

No matter which approach is utilized, it is critical to involve subject matter experts (healthcare professionals in the case of patient engagement chatbots) in such a validation process.

Patient Engagement is Beyond just Q&A

Currently, when considering AI chatbots for patient engagement, AI chatbots are mostly used to answer patient questions. Given the power of AI technologies, AI chatbots can be used beyond just patient Q&A. For example, studies show that cognitive AI chatbots can now infer a person’s personality, such as interests, needs, and temperament, from a conversation. In this light, AI chatbots can engage patients in a genuine two-way conversation, understanding their needs, addressing their concerns, and guiding them to the most appropriate treatment options.

Consider an AI chatbot for mental health. In addition to answering patient questions regarding mental wellness, such as improving sleep quality, it can engage with a patient in an empathetic conversation, automatically inferring the patient’s personality from their conversation and sharing a story that matches the patient’s personality to encourage the patient to maintain their wellbeing. In this case, AI “hallucinations” used to create a story with a fictional character that matches the patient’s inferred personality will be considered valuable because it will make the patient feel being heard and connected through a relatable experience.

Takeaways

Generative AI technologies like those behind ChatGPT are powerful at understanding human languages and composing natural responses. Though hallucinations may occur, they can significantly enhance patient engagements, when used appropriately:

Make healthcare information accessible by customizing AI chatbots with trustworthy healthcare information Further prevent or reduce AI hallucinations by supporting a human-in-the-loop validation process Use AI hallucinations along with other AI technologies (e.g., cognitive AI** to enable deeply personalized, empathetic patient engagements beyond supporting patient Q&A

Note: A version of this article is published on Healthcare IT Today

How can AI like ChatGPT be turned into gold in enterprises?

2023-05-11T21:42:31Z

Generative AI: A Magic Tool for Businesses?

Lately, it's nearly impossible to go a day without encountering headlines about generative AI technologies and their applications like ChatGPT or AutoGPT. AI has become red hot again, and its hotness is astonishing: suddenly almost everyone wants to jump on the AI bandwagon.

As an advocate for the power of large language models (LLMs), I believe that generative AI carries immense potential. In fact, these models have already demonstrated their practical value in enhancing personal productivity. For instance, I have incorporated code generated by the models in my work and even used GPT-4 to proofread this blog post.

The pressing question now is: how can businesses, small or large, which aren't involved in the creation of LLMs, capitalize on the power of generative AI to improve their bottom line?

Unfortunately, there is a chasm between using LLMs for personal productivity gain and monetizing LLMs for business profit. Like developing any business solution, there is much more than meets the eye. Just using the example of creating a chatbot solution with GPT-4, it can easily take months and cost millions of dollars, not to mention the required deep AI expertise and engineering skills.

So organizations must decide whether they want to heavily invest in in-house development or collaborate with a vendor. With many businesses including many enterprises having scarce AI and IT resources, in-house development may not be an option. Their only option would be working with a vendor. But how to evaluate and choose the right one?

The rest of my article will outline the challenges and opportunities to leverage generative AI for business gains, unveiling the lay of the AI land for entrepreneurs, corporate executives, and investors alike looking to unlock the value of generative AI in business context.

Since my focus is on generative AI for business gains, I will start with the business expectations of AI and the valuable lessons learned from AI blunders to better manage such expectations.

Business Expectations of AI

In the modern business landscape, technology is an integral part of business. When a business adopts a technology, it expects the technology to improve its operation efficiency and drive better business outcomes. Businesses expect of AI to do the same, no matter which type of AI is.

Moreover, the success of business does not solely depend on technology. A well-run business will continue to prosper, and a poorly managed one will still struggle, regardless of the emergence of generative AI or tools like ChatGPT.

In general, a successful adoption of AI for business requires that AI MUST perform to deliver concrete business value as expected and the business must know how to manage AI just like managing any other business operations.

Disillusionments of Generative AI

While generative AI has certainly advanced the field of AI especially in natural language understanding (NLU) and natural language generation (NLG), it cannot solve every business problem. Nor will it do. To avoid investing in generative AI hypes, I'd like to first pinpoint out two common disillusionments about generative AI and recognize the associated practical challenges of adopting generative AI for business.

Disillusionment #1

A common disillusionment is that generative AI technologies like GPT or generative AI tools like ChatGPT level the playing field for every business: anyone can use generative AI and English becomes the new programming language.

While this may be true for certain content creation use cases (e.g., marketing copywriting and information summary), generative AI is incapable of solving complex business problems that require deep domain knowledge. For example, ChatGPT generates a medical article with "significant inaccuracies", and it also failed a CFA exam.

While domain experts have the in-depth domain knowledge, they may not be AI savvy or understand the inners of how generative AI works. For example, they may not know how to prompt ChatGPT effectively or refine their prompts to obtain the desired results.

So the challenge is how to enable business domain experts to train and supervise AI but without requiring them to become IT/AI experts.

Disillusionment #2

Another common disillusionment related to generative AI is that any IT personnel can use generative AI technologies like GPT to build awesome business software solutions. While generative AI may help accelerate software development (e.g., writing sample code), all the nitty gritty software development requirements, such as user experience design, data security, system scaling and performance, still need to be met and generative AI's help is rather limited.

This reminds me of the emergence of internal combustion engine, which made car manufacturing possible. While all mechanical engineers or craftsman had access to the engine, not everyone became, nor did every car maker have the success as Ford Motors did.

This also reminds me of the emergence of microprocessor (CPU), which made the making of personal computers (PCs) possible. While virtually all computer engineers and craftsman had access to microprocessors, not everyone became Steve Jobs, nor did every PC maker achieve the success of Apple.

History teaches us while new technology inventions can generate great business opportunities, the opportunities always favor those with deep expertise and passion in the related field: Ford's expertise and passion for mechanics enabled him to capitalize the opportunity brought by internal combustion engine and Job's expertise and obsession with design and user experience enabled him to capitalize the opportunity brought by microprocessor.

As a new technology, generative AI is no exception. Those with deep AI and IT expertise will benefit from generative AI the most and the fastest by capitalizing on the opportunities.

So the challenge is if your organization does not have the expertise to capitalize on the opportunities brought by generative AI, whom should you partner to harness its power?

Valuable Lessons from AI Blunders

The concept of AI has been around since 1960's and the adoption of AI has been through valleys and peaks. One thing agreed among AI experts is that AI is FAR from perfect, including generative AI.

For example, a well-known and challenging issue with generative AI is "hallucination", with AI generating false or harmful output.

Despite rapid improvements in the power and quality of generative AI, advancements in truthfulness have been incremental. While the debate over whether this is a fundamental weakness of generative AI continues among academics, a look at the current state reveals that even the latest state-of-the-art models, like Dromedary, which outperforms GPT-4, can only achieve less than 70% accuracy on the TruthfulQA benchmark. This level of accuracy is not sufficient for most business settings. In other words, no businesses would want to foot the bill for the liability of giving a false response to their customers or prospects 30% of time!

Additionally, there are concerns surrounding privacy and security. For instance, Samsung recently banned the use of ChatGPT due to data leak and Italy also banned the use of ChatGPT due to data privacy.

The experiences and lessons learned from past AI deployments are still applicable and valuable, as they faced similar challenges. Discarding the past efforts to encode domain knowledge and the tools developed to handle complex business cases would be wasteful.

The current wave of generative AI can be likened to a more powerful CPU. If you have already built a computer, you can simply plug in the new CPU to boost performance. This is faster than building a computer from scratch around a new processor. When selecting a vendor, it's helpful to see if they have a track record of success. The decision essentially boils down to whether you trust someone who has never built a computer to utilize the new powerful CPU, or if you prefer someone with prior experience in building computers.

Technical Ingredients for Successful AI Adoption

As the CTO of an AI company, I have witnessed firsthand the challenges and rewards of deploying generative AI in real-world business settings. Based on my experience, I have identified three key areas of technical competence that are crucial for a successful implementation of generative AI solutions.

1. Agent Framework

An agent does not only respond to user requests, it may also take actions proactively on users behalf. An effective agent framework forms the foundation of a generative AI system, allowing it to communicate and interact with users and other systems in a meaningful way. This framework should encompass several components:

Perceptors and actuators: These essential components serve as the AI system's eyes, ears, mouth and hands, enabling it to perceive the environment and take actions accordingly. For instance, perceptors might include natural language processing capabilities, while actuators could involve generating text or controlling other software tools. Generate AI excels in this aspect.
Control Unit: This component should be able to coordinate the perceptors and actuators. A well-designed control unit is vital for managing natural language understanding, generation, and reasoning within the AI system. The power, expressiveness and flexibility of the control unit is critical. There are some general open source implmentation of control unit, such as LangChain, but many use cases require customized control unit.
Working memory: Just like humans, AI systems need a working memory to maintain context during interactions. This component should be able to store and retrieve short-term information, allowing the AI system to respond accurately and consistently to user inputs. In order to process and manage the vast amounts of high-dimensional data generated by AI systems, a vector database is essential for building working memory. This database should be optimized for handling embeddings and other complex data structures, ensuring that the AI system can efficiently access and utilize the information it needs.
Customizable Plugins: Seamless integration with third-party tools and services is crucial for maximizing the AI system's capabilities. By incorporating external resources, the system can expand its functionality and offer a more comprehensive solution to users. These extensions should enable access to customer data and provide additional functionality, ensuring that the AI system remains adaptable and relevant in a variety of use-cases.

2. No-Code Platform

In addition to the large amount of training data, the success of ChatGPT is largely due to the " Reinforcement Learning from Human Feedback" (RLHF), highlighting the importance of incorporating human input in AI solutions. To ensure a seamless collaboration between AI and human expertise, a no-code platform should be implemented.

Human UI: A user-friendly, no-code interface allows business users to interact with and manage the AI system without requiring programming expertise. This empowers non-technical personnel to harness the power of AI, enabling a broader adoption of the technology within the organization.
Test and Evaluation : Comprehensive testing and evaluation processes should be in place to ensure that the AI system meets quality standards and business requirements. Regular evaluations help to identify areas for improvement and drive the ongoing development of the AI system. Tools such as dashboard visualization should be provided to facilitate these processes.
Live human integration: In situations where the AI system is unable to provide a satisfactory response or solution, a smooth integration with live human agents is essential. This ensures that businesses can consistently deliver the desired outcomes and maintain a high level of customer satisfaction.

3. Infrastructure

A robust infrastructure is the backbone of a successful AI deployment, supporting the system's performance, security, and scalability:

DevOps and Scaling : A DevOps approach to AI deployment facilitates streamlined system maintenance, updates, and scaling. By closely integrating development and operations teams, businesses can rapidly adapt to changing demands and ensure that their AI system remains agile and responsive.
Security and Compliance : Ensuring the security and compliance of the AI system is crucial for maintaining trust and confidence among users. Businesses must prioritize safeguarding sensitive data and adhering to industry-specific regulations, demonstrating their commitment to protecting user privacy and upholding ethical AI practices.

Conclusion

As generative AI continues to revolutionize the business landscape, choosing the right AI vendor becomes increasingly crucial for organizations looking to harness the technology's full potential. To make an informed decision, businesses should evaluate potential vendors based on their competence in three critical areas: an effective agent framework, a seamless no-code platform for human-AI collaboration, and a robust infrastructure to support performance, security, and scalability.

By carefully selecting a vendor with proven expertise in these domains, organizations can ensure a successful deployment of generative AI solutions tailored to their specific needs. This, in turn, will help enterprises drive innovation, increase productivity, and gain a competitive edge in an increasingly AI-driven market.

In conclusion, as businesses navigate the rapidly evolving world of generative AI, partnering with a vendor that excels in these three key areas will pave the way for success. By diligently evaluating potential partners and their capabilities, enterprises can make strategic decisions that unlock the power of generative AI and turn it into gold for their organization.

Note

Extensions of this post appear in two articles on VentureBeat and CXSCOOP

How can Juji help you build a chatbot using GPT-4 rapidly with #NOCODE?

2023-05-07T05:46:54Z

How quickly can I build my own chatbot powered by GPT-4 on Juji? What does it take to build one?

In my previous blog, I talked about what it takes to build a chatbot using GPT-4, including the engineering efforts required and a rough cost and time estimate. Although it is non-trivial to build a practical chatbot solution for real-world applications, with the right tools, one can build a custom chatbot solution quickly to take full advantage of the state-of-the-art large language models (LLMs) like GPT-4.

For the past several years, we at Juji have been building Juji Studio, an AI chatbot design tool that enables anyone to build their custom chatbots with #NOCODE. Now with the integration of GPT-4, Juji Studio can enable you to build your custom chatbot even faster and better.

In this blog, I will outline how you can use Juji to quickly build a custom chatbot in each of the five categories mentioned in my last blog. As seen below, you can build these custom chatbots powered by proprietary content and the capabilities of GPT-4 without writing any code or any IT resources.

Type I: Chitchat Chatbots

Although chitchat chatbots are not typical business solutions, you can build them on Juji easily.

Use GPT-4 to Generate Chatbot Messages

Since GPT-4 is integrated with Juji Studio, it can be used to interpret user input and answer user questions whenever appropriate. For example, you can use the basic Juji customer service template (see screenshot below) to create a chitchat chatbot that will let GPT-4 handle user input.

As shown in the example chat below, GPT-4 is used to interpret the user input (i.e., asking for a funny story) as well as produce the response.

In case you don't wish to use GPT-4 to handle user input directly, you can turn it off as shown below.

Deploy Your Chatbot

Once your chatbot is ready, you can quickly deploy it. Your deployed chatbot can be accessed via multiple ways, e.g., via a hyperlink in an email or a chat window installed on a website. Check out this short video to see all three deployment options.

In summary, you can take just 3 simple steps to create and deploy a chatbot on Juji without worrying about web hosting or calling the GPT-4 API.

Select a chatbot template
Preview the chatbot
Deploy the chatbot

Additionally, Juji automatically tracks chatbot behavior (e.g., how many users it has engaged and how long of each engagement) and displays all the information on a dashboard. Below is an example chatbot dashboard.

Type II: Custom Q&A Chatbots

As mentioned in my last blog, Question & Answering (Q&A) is one of the most popular chatbot uses. Juji supports a rapid creation of a custom Q&A chatbot that uses proprietary information. As shown below, the UIUC chatbot uses the proprietary information to answer a user question.

Inject Proprietary Data for Chatbot Q&A

In Juji Studio, you can teach your chatbot with proprietary data in multiple ways. One way is to directly enter a Q&A pair inside a GUI. The other way is to specify all proprietary Q&As in a CSV file and upload the CSV file (training data) into Juji (see below).

You can leverage GPT-4 explicitly to auto-generate alternate question expressions so the chatbot can have a wider coverage for diverse user question expressions. Or you can leave this to Juji, which automatically leverages GPT-4 to capture wider varieties of the given questions. Below is an example of using GPT-4 to generate various question expressions of the question "When will I receive my refund". Again, you still have full control on whether to use the generated questions.

Yet another way to leverage GPT-4 is to simply provide a web page URL or upload other materials to generate Q&As using AI. Once the Q&A pairs are generated, you can validate the generated Q&As and customize them as you see fit. This way, you have made sure that your chatbot will respond to users with the correct information without suffering the hallucinations exhibited by GPT-4 like LLMs.

Since Juji will also host the Q&As for you, you don't need to worry about where to store the data and how to secure and maintain the data either, which again saves you great time and effort.

Q&A Monitor, Report and Update

In a real-world chatbot solution, you'll also need to monitor and update the chatbot so its information is always up to date and its function is improved (e.g., able to answer user questions that it could not answer before). To help you monitor and update your chatbot easily, Juji automatically monitors your chatbot behavior and displays it on a dashboard in real time. By viewing the dashboard, you can learn about your chatbot behavior and update it as needed.

For example, as shown below, you can view how many user questions are answered and how many are unanswered. You can also view various user feedback expressing their thoughts of your chatbot responses. You can then answer the unanswered questions or address user feedback to improve your chatbot instantly without interrupting any existing conversations.

In short, creating a custom Q&A chatbot not only requires language capabilities (NLU+NLG), it also requires engineering efforts to handle the Q&A data, track Q&A status, and the abilities to improve your chatbot whenever needed. The good news is that Juji has already pre-built all such capabilities so you can just concentrate on teaching your chatbot the proprietary Q&A knowledge that it needs to function.

Type III: Custom Interview Chatbots

With GPT-4, Juji enables you to create an interview chatbot easily and quickly. More importantly, Juji still gives you the full control to fine-tune the details of a generated interview chatbot to achieve your specific interview goals.

AI-Generated Interview Chatbots

Just like in a human-driven interview, interview questions may come in different shapes and forms. Thanks to LLMs like GPT, Juji enables you to automatically generate a set of interview questions given an interview context (e.g., interviewing a police officer candidate, onboarding a new student, or gathering user opinions about a product). Below is a screenshot showing how one can specify the interview context.

Using the generated questions, Juji then auto-composes an interview chatbot, which is customizable.

Fully Customizable Interview Flow

The interview flow, in this case, which is also the main chat flow, is fully customizable. The screenshot below shows the auto-generated interview flow of a chatbot that is intended to collect user input about iPhone.

In this flow, you can customize EVERY interview question. For example, you can specify how an interview question should be phrased, whether an interview question is mandatory or optional, or how an interview question should handle a specific type of user response (e.g., negative opinion vs. positive opinion about iPhone). You can also indicate what kind of user input is desired (e.g., it won't accept gibberish input or the user input must contain a minimal number of words). Below are screenshots showing the editable text bubble of a question and how the question could be configured.

Just like in a human-driven interview, interviewees may not always follow the interview flow. In that case, human interviewers must handle out-of-the-flow conversation. The good news is that Juji enables your chatbot to handle such "digressions" or "exceptions" automatically so you don't need to worry that the chatbot might be interrupted and fail at its task. The example conversation below shows how a chatbot handles a digression automatically and resumes the interview flow.

Also note how Ava changed its way of re-asking the same question the second time around. This is because Juji chatbots are powered to recognize a conversation context and always tries their best to make the conversation feel more natural so users are more willing to engage in such a conversation. Again, you don't need to worry about enabling such capabilities since Juji already handles them for you automatically.

Reports and Analytics

To evaluate how well a chatbot has done its interview task, Juji also provides you with a real-time dashboard that contains interview results, such as how many have finished their interview and response analytics on both qualitative and quantitative questions. A csv of organized interview results can also be downloaded for external processing.

In short, creating a custom interview/conversational survey chatbot requires more than just having the chatbot ask user questions and respond to user answers. Such a chatbot must handle specific interview instructions including run-time user digressions. If you wish to create such a chatbot for real-world applications, check out this short series of videos, which will show you step-by-step how you can create a custom interview chatbot on your own with #NOCODE.

Type IV: Custom, Two-way Conversational Chatbots

In many practical business situations, a two-way conversational chatbot is perhaps most useful. Such a chatbot not only can ask questions but also answer user questions. Next, I will show you how to create a two-way conversational chatbot easily on Juji.

Define a Main Chat Flow and Q&A

To enable a chatbot that can support two-way conversations, you only need to define a main chat flow and a Q&A list. Here, the main chat flow as shown above defines a to-do list for your chatbot to carry out, while Q&A list enables your chatbot to answer user questions just like a custom Q&A chatbot. Juji handles the rest automatically, including the interleaving of the chatbot-driven to-dos and the user-driven Q&A. As shown below, one can define a main chat flow and Q&A on the Juji Platform graphical user interface.

Support Multi-turn Q&As

Using the similar method outlined above, you can define a mini chat flow for a chatbot to answer a user's question, which we call a multi-turn Q&A. Using multi-turn Q&As, when a user asks a complex question, the chatbot can break it down into multiple steps to address that question.

As shown below, when the chatbot (Alicia) answers the user's financial aid question, it invokes a flow by guiding the user to navigate through various financial aid options.

In a real-world situation, however, such a question-answering flow may be interrupted. For example, the user may change his/her train of thought and interrupt the current flow by asking "Could you help me find a program first", which is another multi-turn Q&A during which the chatbot will gather certain information from the user before suggesting one or more programs (see screenshot below).

This is known as an interleaving of multi-turn Q&A, where a multi-turn Q&A is interrupted by another multi-turn Q&A. The chatbot must be smart enough to understand which mini-workflow it is in, keep track of each flow, and automatically switch to the active flow (e.g., resuming the last incomplete flow) to ensure its completion. Managing the interleaving of arbitrary multi-turn Q&A could be very tricky and requires sophisticated algorithms given the highly diverse and complex conversation situations. The good news is that Juji's Active Listening dialog engine automatically supports arbitrary interleaving of multi-turn Q&A. So you just need to define a mini-workflow for each specific multi-turn Q&A without worrying about how to manage their potentially intertwined execution.

Multimedia Messages

A lively conversation often contains multimedia messages. Juji enables you to easily create custom multimedia messages that contain images and videos, to be displayed individually or in a carousel.

In summary, a chatbot that can engage users in a two-way conversation is often desired, especially in a business setting. Creating a two-way conversational chatbot is non-trivial as many factors must be considered (e.g., handling the interleaving of chatbot or user initiatives). Nonetheless, Juji has made the process of creating such a chatbot super simple, during which you just need to focus on defining the chatbot's content (e.g., the main to-do list) and let Juji handle the rest.

Type V: Custom, Two-way conversational Chatbots with Third-Party Integration

Third-Party Integration

In case you wish to integrate your chatbot with third-party systems, such as live agent chat or a CRM system, Juji supports third-party API calls within a conversation. Additionally, Juji has built-in integration with several systems, such as Genesys live chat, Facebook Messenger, Google Analytics, FullStory or similar analytics platforms.

Integrate with Live Agent Chat

To facilitate AI-human handoffs, Juji has a built-in Genesys integration. Once you input your Genesys account information (see below screenshot), Juji will take care of the rest.

If desired, you can also customize the transition (e.g., who can terminate the handoff and the transition messages).

Personality Inferences

To personalize each user engagement, Juji chatbots are also powered by Juji's own unique personality inferences. In other words, all Juji chatbots are able to analyze a user's conversation and infer their personality traits in real time. Below shows a user's Big 5 personality derived by the chatbot.

Within a conversation, the personality insights derived by Juji along with the conversation context can be fed to GPT-4 to generate truly personalized messages unique to each user based on their personality. Below shows different personalized messages that persuade someone to quit smoking per their Juji-derived personality.

generated chatbot message for an action-oriented doer

generated chatbot message for a creative and artistic person

As a chatbot owner, you can also view the derived personality report. You can even define a "custom" personality profile that might help you discover your ideal audience (e.g., prospective students who are an ideal match to your program or potential customers who are an ideal match to your service). Below is an example of a derived Big 5 Personality report.

The Importance of NO CODE

While NO CODE enables anyone to build their own custom chatbots, its real power lies in empowering the content owner to fully own the chatbot. Without requiring the IT team to create or manage the chatbot, the domain experts, the marketing specialist, enrollment manager, HR expert alike, can now build, monitor and maintain the chatbot on their own. Much of cross-organizational obstacles, such as time coordinations and communications, can be avoided. The process of creating AI is truly democratized and decentralized.

Start Building Your Chatbot Today

Using Juji Studio, building a chatbot empowered with GPT-4 is no longer a million dollar project. Now you can create a usable and useful chatbot in a matter of hours or days. Additionally, all Juji functions are also accessible through Juji's API in case you wish to embed your chatbot into your existing application (e.g., a mobile app). So why not try it today?

Resources to Get You Started

Sign up on Juji then follow the sign up process.
Easy, no-code customization instructions.
Simple steps for editing Q&A
Step-by-step guide to handle customer Q&A
Juji video tutorials on youtube
How to customize + upgrade Q&As, including multi-step Q&As.
Conversational AI design tips
Why AI skills are important to employees of the future workforce just like computer skills for today's workforce

What does it take to build a chatbot using GPT-4?

2023-04-20T01:03:24Z

Can I build my own chatbot using GPT-4?

How hard is it?

With the popularity of ChatGPT and more recently AutoGPT, more and more organizations (or individuals) are considering the adoption of a chatbot for their website or proprietary applications (e.g., a learning program website or a mobile healthcare application). The pressing question is: should I build or buy one?

The answer is: it really depends on what kind of chatbot you want to have. To determine what kind of chatbot you want, the questions like the following but not limited to the following must be answered:

What kind of tasks is my chatbot set to accomplish?
Whom will the chatbot interact with? How many users will use the chatbot?
What kind of knowledge must it have? Where does the knowledge come from?
Who would maintain or update the chatbot?
What kind of progress report should the chatbot generate and where will the reports be stored?

When I was getting my Master's in Computer Science, I naively thought a chatbot was just a machine learning model that can be set up easily and rapidly. After spending 5+ years on building chatbots for business organizations in their real-world applications, I realized that building a usable and useful chatbot takes much greater effort beyond enabling natural language understanding (NLU) and natural language generation (NLG), or even beyond implementing AI capabilities.

Requirements for Building a Chatbot with GPT-4

Although chatbots can be used in a wide range of applications and industries, their main purpose can be roughly categorized into five types. For each type, I will describe how GPT-4 (which has the latest NLU and NLG capabilities) or an equivalent large language model (LLM) can be used to power a chatbot in that category, including

the simplest implementation, and
additional developments that one might want to consider to make a chatbot useful and usable for real-world applications.

Type I: Chitchat Chatbots

As a student in computer science, this was what I had in mind: a chatbot that entertains you with its humorous sentences on a lonely night (see screenshot below).

If you have access to a LLM like GPT-4, this kind of chatbot is relatively easy to set up. Actually I just asked GPT-4 "What are the steps to create a chatbot using GPT-4?" Below is its answer.

Even though it's wrong about its own release status (GPT-4 is already released!), the main idea was rather straightforward: one can use the GPT-4 API to get input from a user, pass on the user input to get the chatbot response, and then relay the chatbot response back to the user.

This sounds magical: now you could have a humorous chit-chat chatbot on your platform to entertain your visitors. However, what if your chatbot gives out erroneous information just as it did about its own release? What if your chatbot responded to your visitors that other brands offer better products/services than yours? The bad news is that LLMs including GPT-4 hallucinate and might give false statements of fact with confidence. If you are a business owner, you might not want your visitors to get the wrong idea about you or your brand.

Then what if someone tries to attack your chatbot maliciously? Or what if a huge number of users suddenly appeared on your website and overwhelmed your chatbot? You will then need to think about system security and safety, as well as resource scaling just like building any other type of computer applications and systems.

Key Skills Required

Basic Programing
Web Development

Type II: Custom Q&A Chatbots

After reading the above that directly using GPT-4's answers to user questions might not be a good idea for your brand or business, now you want to inject your own information into your chatbot so it can respond to users appropriately. In fact, one of the most common chatbot uses is a custom Q&A chatbot that uses proprietary information to answer user questions. For example, universities want a custom chatbot on their website to answer prospect or student questions about their learning programs (see below an example chat) or hospitals want a chatbot in their mobile app to answer patient questions about their medical services.

Since GPT-4 does not have your proprietary information, you will need to inject the information into the chatbot yourself. While there are many ways to inject knowledge into your chatbot, you might have to make a number of design decisions about your chatbot:

Do you want your chatbot to sound conversational or use the tone consistent with your brand personality when it answers your audiences' questions? Keep in mind, the proprietary data you have (e.g., your website information) may exist in a different form for a different purpose (e.g., long paragraphs or sentences for people to read) instead of in a natural conversational tone. If you care about having your chatbot deliver information in a natural conversational tone, you'll have to transform your data into a conversational one.
What should your chatbot do if it cannot answer a user question? Would you rather have the chatbot acknowledge its ignorance or always provide an answer that might not be accurate? Or do you wish to let the chatbot acknowledge its ignorance for only certain cases but let GPT-4 handle the rest even if its answer might be wrong?
What kind of report do you want to have about your chatbot's Q&A status? Do you want the chatbot to notify you in real time if it cannot answer a user's question? Do you want to use additional channels (e.g., emails or SMS) to notify users that their unanswered questions are now answered? How can you or a chatbot supervisor improve the chatbot quickly if it misses user questions?
Do you want your chatbot to gather user feedback on its answers? How do you want to handle user feedback, especially the negative ones so you can improve your chatbot?

Based on your answers to these questions, you can then decide how to build your chatbot with GPT-4 to handle various situations as described above.

As shown below, the simplest implementation of a custom Q&A chatbot is to include your FAQ list in the system prompt to GPT-4. It will use your FAQ to answer user questions if a user question happens to match one of your FAQs.

However, there are several drawbacks. First, your chatbot may give wrong information to your users when it cannot find any match in your FAQ to the user question. As shown in the screenshot below, my FAQ does not include financial aid information but GPT-4 is not shy about coming up with an answer of its own with confidence, which may pose a liability for you or your business. Second, as the chatbot creator, there is no way for you to be aware of such a situation since this chatbot does not give notifications when it could not find a FAQ match. Of course it won't be able to gather any user feedback either for you to address.

Another more serious limitation of this simple implementation is scalability. As your FAQ list grows, e.g., over a thousand Q&A pairs, you won't be able to upload all your Q&As to GPT-4 since it has a limit on taking in the maximum 8192 tokens (approximately 200 Q&A pairs) at a time.

If you wish to build a custom Q&A chatbot that is scalable and delivers great user experience, things can get even more complex quickly. For example, you'd want to build a real-time chatbot Q&A monitoring system to track when your chatbot fails at answering a user question and then instruct what the chatbot should do in this situation (e.g., routing it to a live agent or taking down a user's contact information for answer notification). To continuously improve your chatbot, you'd also need to implement a system that lets you or other humans update your chatbot with an improved FAQ list without interrupting ongoing conversations. For example, you don't wish to interrupt any ongoing conversations that your chatbot has with a student who is learning about a subject or a patient who is inquiring about medical procedures in order to update the knowledge of your chatbot.

In general, assuming a mid-size simple FAQ list (i.e., a couple of thousands of Q&As with no multi-turn Q&As, see multi-turn Q&A below), plan to budget two full-time engineers (e.g., one working on the backend and one on the front-end), who might be able to build a decent custom Q&A chatbot with GPT-4 in 4-6 months. Keep in mind, this implementation supports only simple, single-turn-based Q&A, excluding multi-turn Q&A that must be supported by the Type IV chatbots described below.

Key Expertise Required

Database
System Design
UI/UX
Web Development

Type III: Custom Interview Chatbots

Another common use of chatbots is conducting interviews to gather information from an intended audience. Studies show that people are more willing to provide higher quality responses in a conversation than in a traditional form-based survey. In fact, organizations have used interview chatbots to conduct social research studies, personality assessments, and job interviews.

Below is an example of interview chat for a police officer position:

To build a useful and usable interview chatbot, you might want to consider a number of design decisions beside determining a list of interview questions for your chatbot to use:

How would you like to control the interview process to ensure the quality of interviewee responses? Can users skip certain questions? Do you want your chatbot to ask follow-up questions or branch off to different sub-flows? Do you wish to reinforce certain constraints on user answers, e.g., the length of their answers? How to handle user irrelevant answers or gibberish input?
What kind of interview report would you like to have? In what format (e.g., CSV vs. text file of dialog history)? Would you like to visualize user answers that might help you identify insights? How would you like to view users' free-text responses to open-ended chatbot questions? Should the text responses be summarized in a way that is easy and fast for human consumption?
How should the chatbot behave when it is interrupted? For example, should the chatbot answer user questions during the interview or just ignore them? What about user's clarification questions regarding an interview question?

As shown below, the simplest implementation of a custom interview chatbot can be made with a system prompt to GPT-4.

Since the main purpose of an interview chatbot is to gather information, it is important to ensure that respondents provide quality input during their interview. The simplest implementation does not allow you to customize how to handle a user's response in various situations. For example, what if a user wants to skip a question, would your chatbot allow it? As shown in the example below, I interrupted the flow with a question, the chatbot answered my question but skipped the current question.

Don't be surprised that human users interrupt a chatbot or an interview flow all the time with diverse expressions. For example, users might give an irrelevant answer to a question, give an excuse to why they won't answer the question, respond with "I don't know", or simply answer with gibberish input. To ensure the quality of user responses, you'd need to implement a full-fledged, configurable interview chatbot to support:

Question configuration (e.g., similar to what other survey platforms have supported the definition of survey questions but more)
Question routing (e.g., if an interview question is interrupted, how the chatbot behaves)
Question communication (e.g., when a chatbot needs to repeat a question, should it repeat the original question or a shorter alternative expression)

It would require tremendous amounts of design and engineering efforts to just figure out how to manage a custom question routing flow intertwined with the GPT-4 chat flow! Expect one full-time AI engineer to spend about 10-12 months to support the functions robustly.

Another important requirement of an interview chatbot is proper recording and reporting of user responses. For practical uses, interview reporting should include the following for the minimum:

Real time recording of interview results
Retrieve and display of interview status (how many ongoing and how many have completed)
Retrieve and display of quantitative responses and analyses
Retrieve and display of qualitative responses and analyses (e.g., qualitative responses require further text processing)

In summary, significant engineering efforts will be required to support interview results recording and reporting, including database setup and management, data retrieval and analyses, and interactive visualization of data analysis results. Budget 1 backend engineer and 1 front-end engineer to implement a decent recording and reporting system in about 4-6 months.

Key Expertise Required

NLP/ML
Data Analytics
Database
System Design
UI/UX
Web Development

Type IV: Custom, Two-way Conversational Chatbots

In real-world applications, chatbots often perform multiple tasks at the same time. For example, a chatbot on a university website may ask visitor questions to get to know more about them while answering their questions about the university programs (see the screenshot below). Likewise, a chatbot inside a healthcare mobile app may ask patients questions to check in on them daily while answering patient questions.

We can consider these chatbots a combination of the Q&A and interview chatbots mentioned above: they can ask as well as answer user questions during a conversation. Because these chatbots are a hybrid of both types of chatbot, many more decision points must be taken into consideration in addition to designing just one type of them:

Who should start the chat, the chatbot or the user, in this two-way conversation? Who should drive the conversation forward? Can each of them drive the conversation as needed? How to customize the configurations?
How to handle arbitrary tasks interleaving? As shown in the example above, a user is asking for a program recommendation. However, before the chatbot could provide the recommendation, the user asks about financial aid.
For complex tasks like daily check-in on a patient or student, how to preserve the conversation context so that a chatbot can engage with a user continuously across a long period of time?
What kind of reporting dashboard would you like to see since chatbots can perform multiple tasks?

Unfortunately, LLMs including GPT-4 does NOT help you address any of the design decisions listed above. This means that you will have to build each part from the ground up. To build such a system, in addition to using LLM like GPT-4 to help chatbot interpret user input or generate certain messages, you'll also need the following:

A front-end system that allows the customization of chatbot tasks and Q&As
Sophisticated algorithms that can handle arbitrary task interleaving
Database management system to track conversation history and preserve context for continued engagements
Chatbot dashboard to display user interaction history or KPIs (e.g., what are the most frequently asked user questions in a given task? Which task has the highest completion rate?)

In general, while GPT-4 could help provide NLU and NLG capabilities, it will take tremendous amounts of engineering efforts to build a practical two-way conversational chatbot that can achieve multiple tasks. Budget 3-5 full-time engineers for about 12-18 months ($2 to $2.5 million) to build a chatbot to support a set of specific tasks. If you wish to support a wide variety of types of tasks, it would require even more effort.

Key Expertise Required

In addition to the expertise required for building a Custom Q&A Chatbot and a Custom Interview Chatbot, building a Custom Two-way Conversational Chatbots also requires the following expertise:

System Architecture
ML Algorithm

Furthermore, to ensure the performance of such system, the following expertise is also desired:

Operation
Distributed System

Type V: Custom, Two-way conversational Chatbots with Third-Party Integration

In an enterprise, chatbots often need to be integrated with other systems to fulfill their tasks. For example, AI is far from perfect, organizations may want to integrate AI chatbots with a live human chat system so users can be transitioned to human agents who can further help them whenever needed (see an example of AI-Human handoff below).

To store and retrieve enterprise data during a chat, chatbots may also need to be integrated with an enterprise's backend data systems (e.g., a CRM system that stores prospect or customer information or a learning system that stores all learning materials).

Each third-party integration will require additional engineering effort, including authentication, testing, system interfacing, and performance scalability. Unfortunately GPT-4 or AutoGPT cannot help you do the integration yet. If you use GPT-4 to build a real-world, two-way chatbot that supports a set of specific tasks and integrate with third-party systems from the ground up, budget 4-6 engineers for 10-12 months of effort ($2.5 to $3 million).

Key Expertise Required

Each third-party integration will require collaboration between both the front-end and backend engineers. The following expertise will be important for integrating third-party systems:

System Design / System Architecture
UI/UX
Web Development

Who's involved in building a chatbot?

Building a chatbot is NOT just an engineering project. For the past few years, I have helped many customers build their chatbots. Most of the time I work with subject matter experts (SMEs) who are none-IT specialists. They are marketing specialists, recruitment gurus, HR experts, and learning specialists. They are the supervisor of a chatbot—setting it up and taking care of the chatbot as it grows. They are also the source of chatbot content since they have the domain knowledge that a chatbot needs to function professionally. For example, a university recruitment chatbot takes content from recruitment specialists, while a healthcare chatbot takes content from healthcare coordinators.

So what I really want to emphasize is the importance of involving and empowering SMEs in your chatbot creation process. This means that when you decide to create a chatbot solution on your own, it is like starting a company-wide project that requires the coordination of multiple departments, involving communications across multiple teams and overcoming organizational cultural barriers and politics. Additionally, such a coordinated effort will have to last forever as long as your chatbot exists because adopting a chatbot is like adopting a child, your chatbot needs continuous nurturing in its existence: to be monitored, maintained and updated. Even with so much effort in place, you might have created just one specific chatbot solution (e.g., aiding students in their learning), which may not be easily extended to handle another use case (e.g., assessing students learning outcomes).

Main Takeaways

In summary, building a chatbot on your own with GPT-4 is possible but requires varied levels of development efforts depending on your goal. Though the simplest solution may be fast and easy to implement, it is often not meant for practical, real-world business applications.

To the minimum, you'll have to

Set up a web hosting platform;
Write a computer program to make API calls to GPT-4 or an equivalent LLM accesses;
Build an interface for your chatbot to chat with your users.

If you want your chatbot to answer or ask questions using your proprietary information, you will then need to support the following minimal requirements:

Integrate with a data storage to ingest proprietary interview questions or Q&A knowledge (simple Q&A only, no multi-turn Q&A as shown above);
Develop algorithms to monitor and control chatbot behavior to handle unknown questions or uncooperative users;
Store, monitor, and report chat results, including chatbot unanswered questions, user responses or feedback.

If you would like to go to the next level with a chatbot that can ask and answer questions in a two-way conversation, the following also need to be supported to the minimum:

Develop a system that can configure, track, and control your chatbot's conversation flow under diverse circumstances (e.g., interleaving of multiple tasks);
Develop a system that can maintain the state of a conversation and handle arbitrary user interruptions properly to manage task completion;
Develop a monitoring, analytics and reporting sub-system that can deliver user insights and chatbot stats for continuous improvement and maintenance.

Furthermore, integration with third-party systems would require another level of efforts:

Develop connections to external third-party systems securely;
Develop supports of customized API calls within conversation;
Develop algorithms to handle complex chat flow management such as human-AI handoff.

Last but not least, keep in mind to involve the content creators all the time, the real AI supervisor of your chatbot. Without them, your chatbot will not be able to get the job done.

What's next?

If you are looking for a better and more efficient way to build a chatbot for your business. Not to worry, help is on the way! See my blog post on how you can use Juji to build a chatbot that uses GPT-4 rapidly with #NOCODE.

Juji Featured in CXSCOOP: New Juji Tool Helps Companies Automate Chatbot Building

2023-03-17T21:42:31Z

CXSCOOP, the leading media featuring latest information technology for customer experience and customer relationship management, published a story about Juji. Senior Tech Reporter, Sandra Radlovacki Vukanovic, recently featured Juji in an article titled "New Juji Tool Helps Companies Automate Chatbot Building".

Our newly launched software allows organizations to swiftly create custom AI-based chatbots without the need for in-depth coding skills or extra IT support. As an extension of our already user-friendly Juji Studio, this innovative no-code chatbot builder marries cognitive AI with an intuitive graphical user interface, giving life to chatbots equipped with advanced 'soft skills'. These skills include closely listening to customer queries and reading between the lines, enabling a truly high-touch and personalized customer engagement.

Our Co-founder and CTO, Dr. Huahai Yang, emphasizes the uniqueness of our AI tool, stating that unlike other platforms, which mainly produce static content, Juji's technology is designed to "generate autonomous AI chatbots that are customized to a specific workflow and proprietary knowledge".

The technology we utilize is not only unique; it's empathetic. Our platform uses computational psychology to equip our chatbots with soft skills that engender more empathetic interactions — a trait that's especially critical when dealing with sensitive topics.

At Juji, we recognize the pivotal role high-touch engagement plays in making high-stakes decisions. This is why our new tool auto-generates AI chatbots custom-fitted to this demand. The chatbots we produce not only engage users proactively but also possess the ability to infer users' personality traits in real-time, ably guiding users through high-stake decision-making processes including, for instance, career choices.

Our Co-founder and CEO, Dr. Michelle Zhou, shed light on the power of generative AI and the importance of harnessing it rightly. "Our new tool enables organizations, with or without AI or IT expertise, to harness the power of generative AI uniquely and safely. This is a significant advancement towards democratizing AI and bridging the AI divide," she remarked.

Indeed, we are thrilled to champion the democratization of AI. Our vision extends to enabling organizations to automate high-value engagements, embedding empathy at every stage of interaction.

For those with a tech-savvy palate, Juji now offers a chatbot scripting solution that provides the capability to put together a chatbot that can handle more complex business tasks like presenting data in interactive charts. This feature also allows for third-party AI plugins or pre-built custom chatbot elements for reuse.

As we celebrate this media recognition, we reaffirm our commitment to revolutionize the AI space and augment high-value, empathetic, and uniquely personalized user engagements. Stay tuned for more updates!

Why AI Chatbots May Be More Empathetic Than Your Boss

2022-09-01T20:40:30Z

If you've interacted with a decent AI chatbot, you've probably noticed that it exudes a certain level of cognitive intelligence – a reflection of human soft skills. This realization initially struck me as amusing as I was poring over chat transcripts with Juji, our chatbot.

During one particular chat, a user inquired about Juji's gender and confessed to having the perilous job of taming tigers without being eaten. In response, Juji noted its lack of being a person and assured the user it would offer help once its capabilities increased. After my laughter subsided, it dawned on me that AI chatbots possess qualities – like humility, transparency, active listening, and the ability to read between the lines – that not only mimic human soft skills but could potentially school us in these areas as well.

Imperfect but Humble

Although flawed, AI chatbots are often designed to maintain a humble tone. They either apologize for any mistakes or confess the absence of certain knowledge. Similarly, leaders who can leave their ego at the door and adopt humility by acknowledging their mistakes and shortcomings can pave the way for a collaborative, empathetic, and harmonious workplace culture.

Transparent Machines

Advocates of AI emphasize the importance of not camouflaging AI as a real person, which often backfires when flaws are noticed, which invariably are, leaving the audience feeling duped. Hence, good AI chatbots are always transparent about their capabilities and limitations. Similarly, if leaders maintain authenticity about their skills, limitations, and are open about their feelings, they can foster an atmosphere of trust, allowing employees to be comfortable being their genuine selves.

The Power of Active Listening

Interestingly, people are more open to machines, especially when divulging sensitive information or brainstorming, because they feel unjudged. Cognitive AI chatbots take this up a notch by actively listening to their users, echoing their sentiments and recognizing their feelings sans judgment. This makes users feel heard and understood. In the same vein, if leaders practice active listening, eschew egos, and direct their focus on comprehending employee's thoughts and feelings, a connection is formed, aiding in more efficient team management.

Reading Between The Lines

Cognitive AI chatbots, much like human psychologists, are equipped to read between the lines by analyzing and inferring a user's implicit needs, interests, and personality traits from their conversation. As the philosopher Cicero opined over 2000 years ago, if one wishes to influence, they must think, feel and speak in the language of the other. Leaders could stand to learn a thing or two from these bot-psychologists when conversing with their employees.

As we stride towards advancing AI technologies, cognitive AI seems to be reflecting our own behaviors by holding up a mirror. Leaders who adopt these humanistic attributes and lead with empathy will go a long way in developing a productive and inclusive workplace. AI and humans, indeed, have a lot to learn from each other.

Note

A version of this post is published as a commentary in Fortune Magzine.

ZDNet Highlights the Unique Capabilities of Juji's AI Chatbot

2022-07-15T11:10:10Z

In the world of AI (Artificial Intelligence) and its ever-growing importance in transforming our daily lives, Juji, Inc. has emerged as a peculiar standout. ZDNet, one of the leading online tech news sites, recently released a thorough coverage of Juji. Delving deep, their feature focuses on Juji's AI chatbot and its distinctive capabilities.

Juji, a California-based startup, has demonstrated a unique approach to AI chatbots. Unlike conventional chatbots, which are mainly scripted and programmed to follow a set of predefined paths, Juji's AI chatbot is built with robust cognitive intelligence. According to ZDNet, this cognitive-powered AI chatbot is designed to mimic human-like conversation effectively and has shown a remarkable understanding of context and sentiment in a digital conversation.

What distinguishes Juji from the pack is its ability to grow smarter with every interaction. By using AI and Machine Learning (ML), Juji’s chatbots learn from every conversation they have, expanding their knowledge and improving their conversational skills. This implies that Juji can understand and remember user preferences, making personalized recommendations or providing tailored responses in subsequent interactions.

But that's not all that placed Juji under ZDNet's spotlight. The AI-driven platform also displays an advanced capability of handling complex dialogues. Whether it’s a debate, negotiation, or simple Q&A, Juji can engage with users in-depth, providing comprehensive and intelligent responses. This feature sets Juji apart from traditional chatbots -rarely seen struggling to maintain the context of prolonged conversations.

The feature in ZDNet also underlined the versatility of Juji's platform. Juji can be utilized in a variety of sectors such as HR (Human Resources) for the screening process, in education for personalized assistance, or in customer service for tackling support ticket overflow. It can lend a helping virtual hand where human resources may fall short or lack efficiency.

Moreover, Juji's AI chatbot offers the delightful attribute of creating fun and engaging conversations. These ‘chit-chat’ discussions are designed to mimic human-like casual conversations rather than solely focusing on problem-solving or information providing. They effectively help to bridge the artificiality gap between machines and humans.

ZDNet’s coverage of Juji, therefore, emphasizes the exciting potential housed by modern AI. In a world receiving AI with open arms, Juji's chatbot offers a new level of sophistication in digital interaction. As ZDNet suggests, Juji's advanced technology could most certainly pave the way for AI’s future, setting new benchmarks for AI-enabled communications.

The technologies used by Juji are more than just promising. They are revolutionary. “Humanizing” AI and bringing it closer to our day-to-day activities, Juji is pushing the boundaries of what AI chatbots can achieve.

In conclusion, ZDNet’s coverage does not only highlight Juji's unique capabilities but also the potential of AI in revolutionizing how businesses interact with their customers and even employees. As AI continues to evolve and grow, it's clear that platforms like Juji will be at the forefront, setting the trend for intelligent, contextual, and personalized conversations. In simple words, the future of digital communication holds exciting possibilities with chatbots like Juji.

Axios Spotlights Juji's Advances in Customer Service Chatbot AI

2022-04-02T10:11:30Z

In her article on Axios, journalist Joann Muller highlights the changing face of customer service chatbots. She underscores that improvements in AI technology have made chatbots more personable and effective in providing assistance.

Over the years, customer service chatbots have become a ubiquitous part of many businesses. Once considered aggravating for their rudimentary responses, chatbots have evolved significantly to offer more person-oriented communication. Muller highlights how new levels of AI have given chatbots the capacity to attribute 'soft skills,' such as empathy, facilitating better communication.

Muller features Bern Elliot, a technology analyst at Gartner, who sheds light on the advancements in machine learning and natural language processing that have made chatbots more adept at understanding and producing language. The chatbots can efficiently interpret a customer's inquiry and analyze the progression of a chat session.

The most successful chatbots are ones that possess human-like personae like Nanci, Sydney, and Erica who represent GM Financial, Anthem, and the Bank of America, respectively.

Though increased chatbot efficiency has led to enormous progress, Muller cites challenges that still exist. These chatbots' responses are programmed to answer from a pre-set library. Thus, the ensuing dialogues can be robotic and the answers limited by the company's data.

As leading tech-enabled giants like Apple and Amazon have demonstrated, the larger the data pool a chatbot has access to, the better its performance in handling customer inquiries.

Taking Axios's coverage a step further, we meet Juji, a start-up that has developed a no-code software that enables even the less technologically inclined to create custom chatbots without writing code.

For instance, the University of Illinois's recruiters used Juji's software to create a chatbot for their high-demand online MBA program. The result was an increasingly conversational bot that improved over time as it had access to more data and could better handle student inquiries.

What sets Juji apart is the software's capacity to discern unique user characteristics, such as their personality type. This allows for tailored interactions. Its chatbots are also empathetic—an unmatched quality in chatbot tech—sometimes even utilizing humor to defuse tension during interactions.

While the trajectory of AI-powered chatbots is certainly on an uptrend, offering remarkable efficiency, their ability to make the customer feel truly understood still leaves room for improvement, as Muller points out. Even with the leaps and strides of AI, humanizing a piece of code enough to understand the intricacies of human communication is a challenge that lies ahead.

Juji's Revolutionary AI Chatbot Builder Featured in the New York Times

2022-03-15T11:12:30Z

We are thrilled to announce that Juji has been featured in a New York Times article titled “'No-Code' Brings the Power of AI to the Masses”. The article by Craig S. Smith explores the rising trend of no-code software that enables anyone to utilize artificial intelligence without needing to understand complex programming.

Our co-founder and CEO, Michelle Zhou, explained our goal of making the development of AI chatbots as simple as creating a PowerPoint presentation. She emphasized our mission to create chatbots embedded with advanced human soft skills, thereby offering a deeper connection and higher interaction quality between chatbots and users.

The article highlights how Juji's no-code interface enabled the University of Illinois to design and manage their custom AI chatbot, thus expanding their student recruitment operations. This achievement showcases Juji's potential to empower organizations to adopt AI and use it skillfully, irrespective of their technical expertise.

The New York Times piece expertly discusses the rise and impact of no-code AI solutions and how they're empowering "citizen developers", people with diverse backgrounds, to harness the capabilities of AI without the need for comprehensive coding skills or technical knowledge.

The no-code movement, which relies on user-friendly web interfaces rather than complex programming languages, is paving the way for the wider adoption and application of AI. It's helping to democratize AI by giving more people the tools to create AI-enabled software and solutions. Juji is proud to be part of this transformation.

But the no-code movement is also encouraging innovation within businesses. As the article notes, most early adopters of no-code AI tools are business professionals seeking to streamline operations and create more efficient workflows without the need to hire programmers or developers. With tools like Juji, they can create custom AI solutions that meet their unique needs.

We are honored to be recognized in The New York Times as contributing to the democratizing of AI technology and look forward to continuing our mission of making complex AI technologies more accessible to users of all skill levels.

As we continue to innovate and grow, we are excited to lead the charge in empowering individuals and organizations to harness the power of AI in accessible and meaningful ways. Check out the full New York Times article here to learn more about the transformative power of no-code AI.

Case Study: Juji Cognitive AI Chatbot Helping Grow Student Recruitment of Online Programs

2022-02-11T21:17:08Z

The Gies College of Business at University of Illinois is a leader in shaping the business education landscape and democratizing business education with a top-ranked, highly successful online MBA program, the iMBA program. The program was launched in 2016 starting with 116 students and has grown rapidly now with over 4600 students. The program has earned a 97% student satisfaction rating, a retention rate of 95%, and the graduates seeing pay raise by 21% on average.

The Challenge

Just like any fast-growing enterprises, one of the main challenges for the iMBA program is to scale its operations to help increasingly more students, both prospective and enrolled students, with limited resources. In particular, the program's limited number of recruitment staff members had to repeatedly answer tens of thousands of questions from prospective students as well as advise them about the program offerings, admission criteria, and the application process. Moreover, there was not an effective way to understand the interests or needs of prospective students, for example, whether they are interested in a degree program or getting a course certificate.

The Solution

To address the challenges mentioned above, Professor Brooke Elliott, the Associate Dean of the College, who oversees the school's online programs, decided to enlist the help of Artificial Intelligence (AI) to augment her team. Professor Elliott stated:

“We have a handful of recruitment staff members who need to answer tens of thousands of questions from prospective students in addition to their other responsibilities such as advising students. A chatbot solution could help us automate certain operations so we can meet increasing student demands while keeping our staff happy."

She wished to explore how AI could automate repetitive work to scale out their fast-growing online programs, satisfying increasing student demands while making staff members happier at the same time.

After watching a quick Juji demo that showed how easy and fast one could set up and maintain a chatbot with no code and no IT resources required, Elliott decided to give Juji a try as she acknowledged:

“this seems easy and our recruitment staff can definitely handle this on their own”.

The Assistant Director of the Admissions and Recruitment team, Myranda Crist, led the team in designing and implementing a chatbot, called Alma, which sits on the program's main web page and performs three tasks:

Answer prospective students' questions about the iMBA program;
Elicit prospective students' information, such as their email and program interest, for follow-up engagements; and
Inform prospective students the key information about the program.

The implementation included two parts, defining the main workflow of Alma as well as teaching Alma how to answer frequently asked student questions. Designing the main workflow of Alma was rather straightforward: the team used Juji Studio, a no-coding AI chatbot design tool, to build the main chat flow of Alma based on a template called AI Program Assistant, which already includes a set of conversation topics related to the three tasks mentioned above and a set of program related FAQs.

As shown in the screenshots below, for example, the first conversation topic (T0) after the welcome message (Welcome) is to invite a prospective student to ask questions and the second topic (T1) is to ask the student if s/he wishes to subscribe to an email list to receive periodical program updates or reminders (e.g., upcoming application deadline).

The screenshot above shows that after sending a welcome message, the AI assistant will invite a website visitor to ask questions (T0). This message also includes a function (help-message-faq) that will automatically generate help messages to inform a visitor what the AI knows. The screenshot below shows what this message looks like in a chat.

If a visitor (prospective student) has done questioning, the AI assistant will ask the visitor whether s/he would like to receive periodical email notifications about the program. As shown in the screenshot below, T1 is the action to ask a student whether s/he would opt-in for email subscription.

To teach Alma to answer prospective students' questions about the program, the team used the Q&A CSV template coming with the AI Program Assistant and prepared a spreadsheet containing frequently asked questions and answers, in a format similar to the one shown below. The spreadsheet was saved to a CSV file format and uploaded into Alma. The screenshot below shows a template of such questions and answers.

As shown above, this is a Q&A template that can be downloaded, filled in, and then uploaded to customize an AI assistant with specific program information.

Alma was designed to enable prospective students to obtain answers to their free-text questions instantly instead of searching through the program website/FAQ list to find answers. As shown in the screenshot below, a visitor can ask free-text questions and obtain direct answers to their questions.

Although Alma could automatically match diverse incoming question expressions (e.g., "What do you know about the online program") with what the staff has taught it (e.g., "Tell me about your program"), Alma was not perfect especially when she was just launched. In the early days of Alma after it was launched, Myranda and her team checked Alma on a near daily basis, reviewing the questions that Alma was unable to answer and adding to Alma’s knowledge base instantly – without interrupting any ongoing conversations with prospective students. Their Q&A list has grown rapidly as unanswered user questions were quickly captured and managed by the team after Alma went live.

Above screenshot is a Q&A board that a chatbot supervisor can check to see what questions a chatbot has missed and can update the chatbot with a proper answer instantly without interrupting any ongoing conversations.

After Alma was deployed, the team also connected Alma with the Google Analytics used to monitor the program website traffic so the team could learn how Alma's interactions with site visitors would impact their website behavior, for example, how many of those who chatted with Alma went on submitting an application.

This screenshot shows how to connect Google Analytics with a deployed AI chatbot to track how user chatbot behavior is related to other website behavior.

The team also used Juji Studio "Reports" dashboard similar to the one shown below to view user chat status and improve Alma as needed.

The Outcome

Alma has interacted with thousands of visitors on the Gies iMBA program website and answered tens of thousands of visitor questions about the online program. From the data gathered, Alma provides several key values.

Improved "conversion rate" by 72%

The key goals for the program website are to inform prospective students about the program, help them find needed information to make a decision, and encourage them to apply for the program. The Google Analytics data of the program website showed that 11.9% of visitors (prospective students) who interacted with Alma the chatbot applied for the program, while only 6.9% of the visitors who didn't chat with Alma applied for the program. This data implies that 72% increase if visitors (prospective students) had chatted with Alma.

Automated 99.5% question answering

Combining Myranda and team's teaching and Alma's continuous learning capabilities, Alma now answers over 99.5% website visitor's questions 24x7x365, covering a wide range of topics about the program, such as admission requirements, tuition and fees, and alumni network information. This relieves much human effort, which could be directed to more important tasks (e.g., content creation and student advising).

Myranda, the leader of the Alma supervising team, mentioned:

"We love Alma. It’s created another way for prospects to connect with us and receive quick answers. Alma has helped inform us of areas in our process that may not be as clear as we would like, and we’ve been able to address that. Another added benefit is that our current students also utilize Alma for answers, something we did not expect!"

Alison Hodina, Alma's direct supervisor, also commented:

"Updating Alma is as simple as responding to an email. I love the ease of Q&A board and how easy it is to teach Alma to answer missed questions, including those with varied question expressions".

Summary

This case study describes the creation and application of a cognitive AI assistant in the form of a chatbot that can multi-task to help grow student recruitment: informing prospective students about the program, automating student free-text inquiries, and eliciting prospective student information.

Adopting an AI chatbot is similar to having a human assistant, a responsible human supervisor should always check in on the chatbot from time to time, identifying mistakes it has made and teaching it new knowledge, especially right after its initial launch, to get the best value of the AI. As the AI chatbot has grown, supervisions could be reduced accordingly (e.g., from daily check to weekly check).

Moreover, cognitive AI chatbots are still a novelty and many people are unaware of their power or have had bad experiences with chatbots that can hardly perform any meaningful tasks. All these perceptions or experiences would deter people from engaging with a powerful, cognitive AI assistant like Alma. We thus strongly encourage organizations who adopt a cognitive AI chatbot to inform their website visitors explicitly what their chatbot can help with and encourage their visitors to engage with the chatbot, which not only will help improve visitor satisfaction but also offload the often unpleasant, repetitive work from human staff members.

Additional Resources

A quick video summarizes the two key steps to create a chatbot like Alma for student recruitment.
How to power a customer service chatbot that is versatile and robust
A recorded webinar on how to create an AI chatbot to greet and guide website visitors (Passcode: 9%Z#K!Z1 to view the recording).

How no-code, reusable AI will bridge the AI divide

2021-12-29T10:12:30Z

In a recently published article in InfoWorld, our esteemed CEO Dr. Michelle Zhou addressed the rapidly evolving landscape of AI technology and discussed a breakthrough that could be a major game-changer across industries - the advent of No-Code Reusable AI. This article provides a profound perspective and insight into the future of AI. Today, we're going to dissect and delve into the key points of the discussion.

Bridging the Divide

The undeniable transformative power of Artificial Intelligence (AI) has swept through the world, redefining business operations and consumer experiences. However, the ubiquitousness of AI's application starkly contrasts with the accessibility of AI technology. There's a significant divide in the AI world with experts enjoying the monopoly of crafting AI solutions on one side while those with minimal tech prowess struggle to navigate it, on the other hand. This gap needs to be closed for us to truly harness the potential of AI.

The advent of No-Code Reusable AI, as our CEO rightly points out, puts the power back into the hands of the general public. The concept of democratizing AI is gradually gaining speed. It aims at enabling and empowering non-technical folks to actively contribute to and influence the AI revolution. By breaking down the barriers of AI usage, no-code AI platforms offer a chance at creating a more equitable and inclusive tech future.

No-Code Reusable AI: Power to the People

The complex code that forms the backbone of AI technology has been deciphered and leveraged by a small niche of programmers and tech enthusiasts to date. The steep learning curve associated with understanding and implementing this code has given it an aura of exclusivity. Enter 'No-Code Reusable AI' - designed to dismantle this very exclusivity.

As the term suggests, 'No-Code' AI implies that anyone, regardless of their knowledge of specific programming languages, can build AI or ML models. It replaces the quaint, complicated code with intuitive, user-friendly interfaces. 'Reusable AI', on the other hand, refers to the ability to use pre-existing AI or ML models built by experts in different scenarios, across different industries thereby saving time, resources, and a whole lot of effort.

So how significant is this?

The impact is seismic. This revolutionary approach opens the door to innovation and ingenuity, fostering a culture of problem-solving and opportunity detection that drives business growth at an unprecedented rate. It eases the transitioning of businesses into the Digital Age and makes AI technology fun, engaging, and significantly less intimidating. Now, more than ever before, people with bright ideas and novel solutions can step into the AI world, learning, growing, and creating groundbreaking solutions.

Conclusion

For long, the power of AI has been restricted to a select few. But as No-Code Reusable AI grows, matures, and evolves, the monopolization is all set to end. This new platform doesn't intend to replace coders but rather, it aspires to bring AI closer to the people, empowering everyone to make the most of this powerful technology. Let's equip ourselves for this fascinating AI future where everyone's a creator, everyone's a contributor, and everyone’s a part of the digital revolution. As our CEO insightfully shared, No-Code Reusable AI is the bridge across the AI divide - let's cross over it together.

An Introduction to Datalevin: An Open Source Contribution by Juji

2021-09-01T05:03:08Z

Juji, an AI-driven conversational technologies company, has made an open-source contribution to the world with Datalevin, a versatile, lightweight, and fast embedded database engine. In addition to being powered by the Datalog query language, it also acts as a Clojure native Key-Value (KV) store. This transactional database solution is agile and dynamic, suitable for handling complex data structures.

Motivation

With the demand for more sophisticated capabilities in conversational AI, there is a need for a Natural Language Processing interface to data. SQL has too many special cases, hence limiting, and XML can be verbose, leaving Datalog as the best target query language. Realizing the importance of granular data access patterns in triple stores led to the development of Datalevin. Datalevin is being utilized by Juji in production for state persistence and full-text search functionalities.

Datalevin Basics

Datalevin's core is a Clojure library, running on Java Virtual Machine (JVM) in an embedded mode. It also have a native command line tool that uses GraalVM native image technology. With various software environments supported, it is capable of running on Linux, MacOS, and Windows. Several features include:

Support as both KV store and Datalog store
Single database file
Sandboxed interpreter for user-defined functions

Datalevin prides itself on being efficient. With the artifact size of only 120KB, it minimizes the dependency footprint, but punches well above its weight in terms of functionality.

Server and Client

With the release of version 0.5.0, the networked server mode was introduced along with the corresponding Clojure client. This major update ensures that Datalevin can transparently work with both local and remote stores, maintaining the same API across both platforms. The server/client mode has these features:

Event-driven server architecture that supports a large volume of concurrent connections
Extensible wire protocol which currently supports nippy and transit encoding
Built-in role-based access control for state-of-the-art security
Built-in connection pool

Full Text Search Capability

Version 0.7.0 saw the addition of full text search functionality to Datalevin. Combining a database with a search engine allows for a more seamless and powerful search across the whole database, thus eliminating the need for incorporating standalone search engines that make the system heavier or redundant. Also, it avoids unnecessary write amplification by storing a reference to the original content instead of duplicating text in the search engine.

The Edge: T-Wand Algorithm and More

The Datalevin search engine is developed using the T-Wand algorithm, which allows it to outperform dominant competitors like Apache Lucene. With the help of this novel algorithm, Datalevin can perform efficient document analysis for the given query, even in situations that demand a high level of complexity.

Conclusion

In conclusion, Datalevin clearly demonstrates that a versatile transactional database does not need a large number of dependencies and a bloated code base to deliver the best performance. Experience Datalevin today for flexible Datalog query and good performance, brought to you by the innovators at Juji, Inc.

Juji, Inc. Powers Cornell University Study on Human-Machine Team Creativity

2021-07-27T05:33:33Z

Study finds when participants perceived their teammate to be a chatbot, they consistently generated more, higher quality creative ideas

SARATOGA, Calif. — June 29, 2021 — Today, Juji Inc., an Artificial Intelligence (AI) company that specializes in developing cognitive AI assistants, announced the findings of a study, “IdeaBot: Investigating Social Facilitation in Human-Machine Team Creativity,” by researchers at Cornell University’s College of Agriculture and Life Sciences (Cornell CALS). The study utilized Juji’s AI platform to investigate how humans collaborate with a cognitive AI assistant, in the form of a chatbot for a creative brainstorm.

The study found the following:

When human participants perceived their teammate was the Juji chatbot when participating in a creative idea task, they consistently produced more, higher-quality ideas.
When the conversational style of the chatbot was robotic (versus human-like), participants with high anxiety in group communication reported greater creative confidence.

“When working with the Juji chatbot, human participants worried less about judgments from their partners, and they yielded more fruitful conversational outcomes in the task,” said Angel Hwang, the study’s co-author and a graduate student working with Andrea Stevenson Won, assistant professor of communication and director of the Virtual Embodiment Lab at Cornell CALS. “This research supports the promising role of autonomous agents like chatbots as active teammates and points to ways in which leveraging the distinction between artificial and human cues during conversation can drive positive outcomes.”

Hwang and Stevenson Won utilized Juji to design the conversational flow, customized certain functionalities of the chatbot to follow the conversational script written for three studies and designed the language used in the conversation with specific emphasis on being polite and formal.

“We are thrilled to provide the researchers at Cornell CALS with our Juji cognitive AI assistant platform for the study,” said Michelle Zhou, CEO of Juji. “The findings showed that not only do chatbots powered by cognitive intelligence, like Juji’s, encourage greater participation and comfort for users, but can take on a meaningful role in collaborative, team settings. This can have far-reaching implications for not only universities, but businesses across every industry.”

Read the full announcement.

4 Must-Haves for Universities When Considering a Chatbot

2021-05-18T22:10:05Z

Companies around the world are constantly evaluating how to best reach and/or serve their customers, exploring ways to better connect with customers, while always keeping a keen eye on the resources needed to do so. Higher education is no different. It too needs to cut through the clutter and create a connection with potential students, and continue to build the relationship even after those students are part of their university environment.

As the global chatbot market continues to expand (according to a report published in February 2021 by Research Dive), many brands including universities are now incorporating chatbots as part of their admissions teams. While more universities incorporate chatbots as an extension of their workforce, they also often find typical chatbot solutions are inadequate, leaving their users frustrated with an unsatisfactory, impersonal experience. The personal connection becomes lost with chatbots ignoring user inquiries, offering irrelevant or often, one-size-fits-all responses. Herein lies the challenge — can a chatbot automate human conversation AND be personal?

Today a new generation of chatbots, called cognitive AI assistants, can do just that. Additionally, organizations can launch them in a matter of days vs. months. The incorporation of cognitive intelligence into a chatbot immediately transforms users' experience. Instead of being forced to follow a fixed interaction path (e.g.,clicking on menu buttons one after another), a user is now interacting with a chatbot that has a more human-like quality than typical chatbots, and that can learn, remember, reason, solve problems and make sound judgements. How would universities tell a cognitive AI assistant from a typical chatbot? A cognitive AI assistant has three key attributes that universities should look for.

Responsible AI

When universities evaluate a chatbot solution, it is important to verify whether such a chatbot is adequately responsible to complete its assigned tasks, whether it is to help prospective students obtain program information or existing students access course information. A responsible AI assistant can truly help universities help their prospective and existing students 24x7x365.

Below shows two conversations. On the left (a), it is a typical chatbot that can hardly understand how to help its users, while the chatbot on the right (b) is a cognitive AI Assistant that answers a user's question promptly and maintains the conversation context to ensure task completion (in this case continuing the online application process).

Empathetic AI

In addition to evaluating a chatbot's abilities to accomplish tasks, when universities evaluate a chatbot solution, it is also important to verify whether such a chatbot can help their brand build empathetic and deep personal relationships with their prospective and existing students.

Due to a lack of intelligence, a typical chatbot can hardly understand its users, let alone inferring users' characteristics. That's why a typical chatbot often gives a one-size-fits-all, impersonal response regardless of individuals' needs and wants. In contrast, a cognitive AI assistant can read between the lines during a conversation and automatically infer a user's unique characteristics, such as the user's interests and preferences. Discovering a person’s “internal state” from the person's conversational text, a cognitive AI assistant enables universities to deeply understand and better help their students (prospective and existing alike), as well as create deeply personal connections with them in real time.

Below shows how a cognitive AI assistant personalizes its guidance to two different users based on its understanding of the users' different characteristics.

Real-Time AI Facilitating Human-Chatbot Collaboration

Since no AI is perfect and also needs improvements, when universities evaluate a chatbot solution, it is important to examine how such a chatbot would work with their human teams, which must monitor and update the chatbot from time to time. This is also because universities not only want a chatbot to be an extension of their brand, but also a valued member of the team to elevate the overall work environment.

Unlike a typical chatbot that is often unaware of user-chatbot interaction situations or is hard to be improved, a cognitive AI assistant can work collaboratively with its human teammates, informing the teammates about its own status and suggesting ways to improve itself. Using such information, the human teammates can improve a cognitive AI assistant instantly without interrupting any ongoing conversations. The following example shows how a cognitive AI assistant informs its human teammates about its inference of a user (Mary) and her unanswered question. Given such information, the human teammates can address the unanswered question while using the information to improve the AI immediately.

![The image shows a user profile derived from a conversation and the chat transcript between the user and the chatbot. ](/assets/uploads/screen-shot-2021-05-18-at-3.52.35-pm.png "A \"cheat sheet\" generated by a cognitive AI assistant based on its interaction with a user. Such information can be used to aid human agents in following up interactions with the user.")

Cost and Speed of AI Deployment

No university has unlimited resources and adopting an AI solution, especially a solution as powerful as a cognitive AI assistant, can be very costly. When universities evaluate a chatbot solution, it is also important for them to evaluate the cost of incorporating such an AI assistant into the “team”. There are normally two types of cost involved in implementing such a solution: human resource and time required. Historically, it would take months to set up and deploy such a solution, while new advances in AI enable rapid setup and deployment, in a matter of days. Moreover, it is important to ask what kind of human resource would be required to implement a cognitive AI assistant solution? Will it require AI expertise to set one up? Will there be a need for dedicated IT support? Who will build the situational experiences into the chatbot? As most universities have stressed resources, finding a solution that does not require a dedicated IT resource or technical staff is vital for the quick launch of a cognitive AI assistant.

Solutions that have rich, pre-built AI capabilities (e.g., AI templates or AI training tools) will cut the implementation time tremendously. In turn, they allow universities to start taking advantage of a cognitive AI assistant faster and perform ongoing maintenance with little effort.

A shorter version of this guide was also published by University Business, a media outlet that provides coverage of education news, technology, academics, facilities management, security, financial services, policy, profiles, and opinion.

Juji Featured on Cheddar TV: Understanding Cognitive Intelligence

2021-05-13T23:24:11Z

Juji Co-Founder and CEO, Michelle Zhou, talked with Cheddar TV on Juji's work on cognitive intelligence and the state of AI chatbots.

The discussion is around three questions.

WHAT'S CURRENTLY MISSING WHEN IT COMES TO A.I. CHATBOTS NOW, AND HOW IS JUJI CHANGING THAT?

When it comes to existing AI chatbots, two things are often missing.

First, most AI chatbots actually has little AI in them. For example, a typical chatbot often forces its users to click on a bunch of buttons and follow a fixed path that often leads nowhere. Additionally, no matter who interacts with the chatbot, and the chatbot always gives one-size-fits-all responses.

Second, it’s expensive to acquire AI chatbots. It requires AI expertise, programming skills, not mentioning precious resources, such as AI training data.

What Juji brings to the table is a new generation of chatbots: not only do they have AI, but they also have an advanced version of AI--human-like cognitive intelligence. We call them cognitive AI assistants. Juji also enables almost anyone to quickly set up, launch, and manage their custom cognitive AI chatbot. No coding is required and no technical skills are required.

HOW DO YOU DEFINE 'COGNITIVE INTELLIGENCE'? WHAT ROLE DOES IT PLAY IN THE CHATBOT INDUSTRY?

In our human world, "cognitive intelligence" refers to our human abilities to understand, remember, learn, complete tasks, and make decisions. In the machine world, "cognitive intelligence" refers to machines that possess the very similar, human-like intelligence with their capacity to understand, remember, learn, and complete tasks.

Because of human-like intelligence, Cognitive AI can fundamentally transform the chatbot industry from three aspects:

First, chatbots are no longer just dumb bots. Instead they can augment a workforce to serve as humans’ responsible assistants in a wide range of tasks, from providing customer support to interviewing job candidates to conducting market research.

#2 Chatbots are no longer just cold, impersonal software. Just like humans, now they can have social-emotional intelligence, deeply understanding their users, empathizing with the users, and personalizing each engagement. For example, in the context of helping the higher education sector, a Juji AI chatbot can be configured to offer deeply personalized learning assistance or career development information.

#3 AI chatbot is no longer a luxury solution afforded only by certain organizations. Because Juji enables almost any organization to create and manage their own cognitive AI assistant, without requiring additional technical resources.

WHY IS IT IMPORTANT FOR YOUR USERS TO KNOW THEY ARE TALKING TO A BOT?

When adopting AI, it is critical to be transparent with AI users and let users know that they are interacting with a machine not a real person for two reasons.

First, AI is far from perfect. An organization needs to set the right expectations for their audience when using AI to engage with the audience. This way users know that they are interacting with a machine and don’t expect them to behave perfectly. People are most likely willing to forgive AI mistakes but they may not be so forgiving if the AI pretends to be a real person.

Because AI is imperfect, people often treat AI as a child and they are more honest and true to themselves when they are interacting with an AI as if they are interacting with a child. This is because they believe the machine or a child won't judge them like an adult person would do. This helps create an authentic relationship between a user and an AI, which helps the AI better understand and help the user in his/her true self. Just imagine that if AI tries help a student to find a major or a career development path, the authenticity of the student would help AI make better suggestions that suit the student and the student's true characteristics.

The full interview can be viewed online.

Juji Featured in EDSCOOP: AI Chatbot Maker Juji Jumps into Higher Education

2021-05-03T21:42:31Z

EDSCOOP, the leading media featuring latest information technology for higher education, published a story about Juji and Juji's cognitive AI assistants for the higher education sector.

The story mentioned "Juji’s stand-out feature is its conversational flexibility, a design approach rooted in decades of linguistic and AI research. While it’s common for commercial chatbots to get stuck when confronted with an uncooperative or confused interlocutor, Juji employs human-centered design, ... politely responding to off-topic input, but not losing sight of the interaction’s purpose."

The story also highlighted a potential use case of Juji cognitive AI assistants for helping university counselors, "Chatbots can collect patient (student) history before an appointment, saving valuable session time and providing the counselor with a “cheat sheet” of what’s bothering the student."

About EDSCOOP

EDSCOOP is the leading media brand in the higher education IT market. With more than 1.2M monthly unique engagements, EdScoop features the latest news and conversations among the top leaders driving technology and digital learning in higher ed. With our website, newsletter and events, EdScoop uniquely attracts the most influential IT decision makers at universities around the world.

Juji, Inc. Unveils Cognitive AI Assistants for Higher Education

2021-04-21T21:01:17Z

SARATOGA, CA - April 22, 2021 - Juji issues a press release and announces its offering of a new generation of chatbots that makes AI accessible to universities and brings human-like engagement to students without requiring IT resources.

As universities grapple with how to bring greater efficiency to their operations such as the admissions process, and expand their workforce coming out of the pandemic, today, Juji Inc., an Artificial Intelligence (AI) company that specializes in developing cutting-edge AI solutions, unveiled its cognitive AI assistants for the higher education market.

“It’s challenging for universities to keep pace with thousands of inquiries coming in from current and prospective students and alumni, particularly at a time when resources are tight,” said Dr. Michelle Zhou, CEO of Juji and an expert in human-centered AI. “Cognitive AI assistants are the new generation of chatbots that will empower universities and other organizations to respond to incoming inquiries 24x7 and personalize these responses whenever possible.”

Unlike other rudimentary chatbots being deployed at universities that often leave their users with a frustrated experience, Juji’s cognitive AI assistants engage users in a more human-like interaction by actively listening to the users and automatically inferring their unique needs and wants. Juji AI assistants are able to remember, learn, and solve problems, enabling them to act responsibly and empathetically—more deeply understand and better help users.

Juji also enables universities to set up and deploy custom Juji AI assistants within days with no coding or technical resources required. This recent University Business story informs how universities should evaluate chatbot solutions to maximize the chatbot ROI. A free consultation with Juji can be scheduled here.

Dr. Zhou continued, “We are thrilled to be making AI accessible to universities and education institutions. Cognitive AI assistants are empowering universities to quickly respond to questions while building personal relationships with their students and alumni, streamlining operations, and helping increase enrollment.”

ABOUT JUJI INC.

Juji, Inc is a leading AI company that specializes in powering a new generation of AI assistants with cognitive intelligence. With headquarters in Silicon Valley, the company was co-founded by Dr. Michelle Zhou and Dr. Huahai Yang, inventors of the IBM Watson Personality Insights and the world’s top experts in Human-Centered AI. Juji enables non-IT personnel to rapidly build and launch a custom AI assistant using an intuitive graphical user interface in a matter of days. For more information, visit https://juji.io/.

Search Engine Journal Podcast: How Cognitive Intelligence will Reshape Chatbots

2021-04-14T23:53:43Z

In his podcast, Brent Csutoras at Search Engine Journal had a conversation with Dr. Michelle Zhou, co-founder and CEO of Juji on how cognitive intelligence will reshape the chatbot industry.

"We have heard for years that Chatbots are the future and although we have seen some significant progress in the concept and application, they are still impersonal and not as helpful as we had hoped they would be."

In response to Brent's question, Michelle talked about what "cognitive intelligence" means in the AI world and how it enables chatbots to be much more effective--being both empathetic and responsible in their tasks to truly understand and help their users.

Check out the full podcast for more details.

Case Study: Accelerated Product Insights from a Customer Service Chatbot

2020-10-27T04:06:02Z

JumpStart® is the leader in creating interactive experiences that enrich, entertain and educate. It produces high-quality educational games that provide positive, safe and fun experiences. Their games have earned the trust of millions of teachers, parents, and respected organizations such as Common Sense Media and The National Parenting Center.

The Challenges

As online learning has become increasingly more important, the JumpStart team developed and launched a new online service called Activity Studio. This service enables teachers to create, share, and incorporate fun, game-based learning in their teaching activities. Additionally, these online learning materials can help improve student engagements and boost students' learning interests.

When Activity Studio was launched several months ago as a new service, the team faced two challenges. First, it was time consuming and labor intensive to manually respond to numerous user questions because users often prefer obtaining direct answers instead of wading through a website to search for potential answers. Second, there were no natural and easy ways to engage with users and help them out while they encounter difficulties (e.g., helping users when they abandoned certain activities), as well as gather their real-time feedback to glean product insights and use such insights to improve the service.

The Implementation

To address the two challenges mentioned above, the service team designed and implemented a chatbot, called Johnny, which performs three tasks:

Anwer users' questions about the service
Elicit users' input
Teach users key service features

There are two versions of Johnny, one serving teachers while the other serving students.

Anwer free-text user questions about the service

Like any new service deployed, initial users (i.e., teachers and students) normally have questions regarding the service. For example, the users want to know how the service can help them or a particular feature of the services. Since such user questions can be quite diverse, Jonny was designed to enable users to obtain answers to their free-text questions instantly instead of finding answers by searching through a website/FAQ list or being forced to traverse down a fixed tree-based menu.

As shown in the screenshot below, a user can ask free-text questions, including questions that require multi-turn interactions.

Elicit User Input

While Jonny helps answer user questions and solve user issues, it was also designed to ask questions and elicit users' input. For example, Johnny asks teachers which grade and subject they are teaching and their NPS score about Activity Studio (see screenshot below).

This chatbot behavior is quite similar to human-driven call center services. When a customer calls a call center for support, the customer may be asked to fill in a survey to provide his/her feedback about the service received.

Teach Users Key Features of New Service

In addition to answering and asking user questions, the chatbot Johnny also serves as a "service guide" to teach users the key features of Activity Studio. As shown below, it informs teachers about the most frequently used service features and encourage the teachers to check them out.

Use Juji Studio and Juji API

To support all three functions described above, the JumpStart team uses Juji Studio, a no-code AI chatbot builder, to build the main chat flow of Johnny. The team also defined and uploaded a small Q&A list to start with, including the most common questions, such as "What's Activity Studio". This Q&A list has grown rapidly as unanswered user questions were automatically captured and managed after Johnny went live.

Although the JumpStart team can directly embed Johnny into their website, the team decided to use Juji API to integrate Johnny into the Activity Studio web application. This enabled Johnny to be an integral part of the website (e.g., Johnny could be activated based on user activities on the website) and to appear in a look that is consistent with the rest of the website.

After Johnny is deployed, the team also used the Juji results dashboard to monitor Johnny's status and improve Johnny quickly.

The Outcome

Johnny has served over 50,000 users during the last six months. It provides several key values as described below.

Automate Customer Support

Johnny automated frequently asked customer questions (e.g., answered thousands of user unique questions with 73% answering rate), such as the use of certain service features (e.g., how to share a game) and certain operations (e.g., login), 24x7. This relieves much human effort, which could be directed to more important tasks (e.g., content creation and feature improvements). In the meantime, this also improves user experience and satisfaction because they could get a direct answer they want anytime without waiting for a reply or searching for the website.

Product Insights

Through the interaction, Johnny was also able to learn what features users care about the most as well as their needs and pain points. For example, teachers' questions, such as "how can I view my students' results" and "how do I check my students' learning progress" , enabled the product team to learn user feedback and improve the reporting functions of Activity Studio to make the service truly customer centric.

"Activity Studio has been evolving rapidly and improved constantly due to the fast customer feedback gathered by Johnny. Without the chatbot, I could not think of a more natural and better way for us to engage with our users and collect their feedback to improve our service so quickly." stated by Tracy Chi, who is the Senior Director in charge of the development of Activity Studio.

A Natural Channel for Introducing Product Features

Johnny is also a natural channel for the product team to introduce new or important features to users. Not only does this help the product team learn about user needs (e.g., users care about one feature other the other), but it also enables users to learn about the service easily, saving users' time to search for needed features or services.

Summary

In short, this case study describes the creation of an online learning chatbot that performs three functions in one: automating customer services (free-text Q&A), eliciting product insights, and introducing product features. It also demonstrated a successful use of a customer service chatbot that can multitask to serve the purposes of product development and product marketing at the same time.

Additional Resource

How to create a multipurpose chatbot that can perform multiple tasks, including customer service and marketing in one chat.
How to power a customer service chatbot that is versatile and robust
Sign up to create your own multitasking customer service chatbot and automate customer support as well as product marketing and product insights gathering.

Reading Between the Lines: A Guide to Chatbot Personalization by Personality

2020-10-19T20:30:38Z

As indicated by this post, 84% consumers want to be treated as a unique individual not just a number, while 95% of companies saw 3X ROI from their personalization efforts. Now with the rise of conversational commerce, chatbots become a natural and private channel for brands to engage with users and offer personalized services in one-on-one conversations. In such a conversation, a chatbot can naturally elicit users' needs and wants and then provide personalized help or guidance based on the gathered information.

However, most chatbot personalization has been done based on users' explicit information, such as user past activities (e.g., purchasing history) and user explicit preferences (e.g., "I want a car under $5000"). While such information can help personalize a chatbot's interaction with a user, it is often inadequate in many real-world applications.

Chatbot Personalization Examples

Consider the following examples that cover a wide variety of chatbot applications, including e-commerce and healthcare. To serve a user as a unique individual, these examples show that a chatbot must know a user much more deeply beyond his/her past activities or explicit preferences.

Personalizing Product Recommendation

Sam is a new customer to a restaurant and has never ordered anything from the restaurant before. Although Sam explicitly expressed that he prefers fish to red meat, there is still a wide range of selections. What kind of dishes should the restaurant's chatbot recommend to a new customer like Sam? If the chatbot knows that Sam is family-oriented and enjoys down-to-earth food, it could then suggest the homestyle, fish banquet to Sam and his family.

Luna just got her driver's license and wants to buy her first car. Since Luna has never owned a car before, she could provide only a price constraint, e.g., under $15,000. With a number of cars that meet the price constraint, what kind of cars a chatbot on a website like carmax should recommend to Luna ? If the chatbot knows that Luna has keen artistic interests and loves stylish products, it would recommend a stylish car like Mini Cooper or VW Beetle.

Personalizing Wellness/Healthcare Nudging

Emma uses a wellness mobile app for anger management. Emma is supposed to interact with the app, document her emotional states, and practice anger management on a daily basis. How should the app's chatbot encourage Emma to stay on course when Emma does not do what she is supposed to do? If the chatbot knows that Emma is ambitious and values self-empowerment, it can encourage Emma to stick to the program for the benefits that she cares about.

Amy has diabetes but loves sweets. How should a chatbot that serves as her personal wellness assistant can help persuade Amy to reduce her sweets intake? If the chatbot knows that Amy is intellectually curious and like challenges and winning, the chatbot may suggest that Amy solves a puzzle whenever she craves sweets.

Personalized Career Advice

Phil is a senior in college. To prepare for his upcoming job interviews, Phil practices interviews with an interview chatbot. What kind of advices the chatbot should give Phil and help him succeed in his job interviews? Knowing that Phil is thoughtful and cautious but often unwilling to state his opinions, the chatbot can encourage Phil to be bolder and more decisive when expressing his thoughts during an interview.

Challenges in Learning Personality

All the examples above show that a chatbot would be able to offer personalized suggestions and advices if it knows the personality of its users. Broadly speaking, personality includes a set of characteristics that define a person as a unique individual, also known as individual differences. One's personality distinguishes a person from another from three broad aspects:

What a person likes to do, including the person's passions and interests.
What a person is good at, such as the person's talents and skills.
How a person deals with life's challenges, such as handling emotional and social challenges.

Now comes the question: how can a chatbot learn each user's personality? Traditionally, psychologists learn individual differences by asking people to take item-based surveys, such as a personality test. In such a survey or test, people score themselves on a set of items. A chatbot can certainly use the same approach to learn its users' personality by asking each user to take an item-based test.

However, such an approach has several drawbacks. First, it may be awkward to ask a user to take a test (e.g., "Welcome to Walmart, would you like to take a personality test?"). Second, users may not be always truthful in such a test due to social desirability biases especially in high-stakes situations (see testers cheat on the item-based self assessment). Third, the items in a test are usually come up by Psychologists in a lab, and may be out of context for some users. For example, if a person has never watched ballet, the test item asking him/her how much s/he likes ballet would be out of context. Fourth, it is difficult to assess the results objectively based on self scoring. For example, on the item "I love ballet", one person's self-score of 5 might mean a different degree of love than that of another person. Additionally, people's personality does change over time, it would be burdensome and impractical to ask users to take tests repeatedly and track their changes.

To overcome these challenges, computational approaches have been proposed. One popular approach is to ask a group of users to take an item-based survey and then collect the users' behavior data (e.g., communication text). Regression models are then used to examine which features of user behavior (e.g., words or word categories) could be used to predict survey scores. Such approaches have been used to predict Big 5 personality scores, values, and needs. However, this approach is expensive and still relies on item-based survey results to come up with computational models.

Reading Between the Lines

In the search of more efficient and objective approaches to measuring individual differences, Juji has invented a new model that combines big data analytics with Item Response Theory (also known as latent trait theory) to automatically infer individual differences from user behavior. Currently, Juji uses this approach to infer Big 5 personality scores, including all the 35 dimensions, from text. Specifically, with this model a chatbot can interact with users in a natural way without asking them to take a Big 5 personality test. It analyzes a user's conversation text on the fly to infer the user's individual differences. The chatbot can then use the insights to personalize each conversation as shown by the examples above.

If a user interacts with such a chatbot continuously (e.g., asking the chatbot questions), the chatbot can automatically learn the changes in the user's individual differences in a natural way and use such information to better help the user (e.g., offering suitable career development advice).

Juji currently focuses on inferring the Big 5 personality traits as its base model, because numerous research studies show that these traits characterize the fundamental individual differences in people and they can influence people's real-world behavior and predict important life outcomes. On top of these basic traits, Juji infers context-specific user characteristics, e.g., one's reader DNA, which can then be used to help users find products or services that match with their innate needs. For example, a chatbot would recommend self-enhancement or business books to cerebral readers, while recommend art-related books to aesthetic readers.

Reliability: How Stable Are the Inference Results

To assess the quality of the trait inference results, it is important to measure the reliability of the method. Reliability is to assess how stable the inferred scores are under consistent conditions. For example, if given a set of 100 randomly sampled words produced by a person, how do the inferred trait scores differ for each set of words? The image below shows the computed Cronbach's alpha for the inferred scores of the six facets under the Extroversion dimension in the Big 5 personality model.

The above graph shows that the model produced an acceptable reliability (> 0.8) for all the facets except for Cheerfulness, if the number of words used in the inference was over 1000. This means that the inference of those facets is reliable if it uses more than 1000 words.

Validity: How Useful Are the Inference Results

In addition to reliability, another important measure is validity, which assesses whether the inferred trait scores correspond to real-world user behavior (e.g., predicting a user's behavior or preference). Since measuring validity is often context sensitive, experiments or studies need to be carried out in a particular context. For example, Xiao et al. show that personality trait scores inferred by a Juji chatbot can predict team performance.

Chatbot Personalization by Personality

To help chatbot designers easily personalize each conversation experience, Juji provides a set of functions and "message cards", which allow a designer to personalize chatbot messages for each inferred user persona.

How to Personalize Chatbot Messages

One easy way on Juji to personalize chatbot messages by individual differences is to customize a pre-made "message card". Currently Juji provides two pre-made message cards: one allowing a chatbot designer to customize chatbot messages to match inferred shopper personas, while the other allowing a chatbot designer to customize chatbot messages to match a set of base personas.

As shown in the screenshot below, a chatbot designer can personalize an e-commerce chatbot's messages by different shopper personas. In a live conversation, a chatbot will send a custom message to a user based on the user's shopper persona inferred from the conversation.

The screenshot below lists all shopper personas that Juji automatically infers. In other words, a chatbot can send different messages or recommend different products to users with different shopper personas. For example, a chatbot may recommend high-quality, brand-name products to users with an inferred "Achiever" persona, while suggest innovative, new products to users with an inferred "Explorer" persona. To ensure accuracy, currently a chatbot would not send out a custom message if it does not have enough information (e.g., few words) to infer a user's shopper persona.

Likewise, a designer can personalize a chatbot's messages to different users based on their base personas, such as giving career advices. Again, such a chatbot just needs to carry on a natural conversation with a user, such as asking users how they are doing, and use such a conversation to infer the user's individual differences and then send custom messages based on the inference results. As shown below, Juji also automatically infers a set of base personas.

A chatbot designer can easily customize a chatbot's messages to users based on their inferred base personas. For example, a designer can customize a chatbot's message to one who has a "thinker" persona (thoughtful and cautious) versus to a user who has a "helper" persona (warm-hearted and caring).

In addition to using pre-made message cards as shown above, one can use system functions as preconditions to personalize chatbot messages by individual differences. For example, if you wish that a chatbot sends a message only to someone who is caring and warm-hearted (is-helper), you can then insert this function as a precondition of a remark so the message will be sent to only users who possess such characteristics. See below the screenshot.

When to Personalize Chatbot Messages

Now I have explained how a chatbot designer can easily customize a Juji chatbot's messages to address different user personas, the next question is when a chatbot designer should do so to get the best value?

Help Cold Start

One effective use of chatbot personalization by personality is to help cold start, such as onboarding a new user or introducing a new product. In such a situation, a chatbot often does not have other sources of information to personalize its messages. For example, a chatbot does not have the past activities of a new user or it does not have user purchasing history or reviews of a new product.

For a new user, a Juji chatbot can chat with a new user by asking the typical on-boarding questions. In addition to gathering the basic on-boarding information, this chatbot will also automatically infer the user's individual differences from the conversation and now can make recommendations without the user past activities.

For a new product, a Juji chatbot can describe the value of the product that matches the personality of users even though the users have not experienced the product yet. For example, users who desire safety would love safe, high-quality products, while users who are highly curious and open to new experiences would enjoy innovative products.

Help Longitudinal Engagement

Knowing users' individual differences also helps a chatbot engage with users longitudinally. For example, a chatbot serving as a personal wellness assistant will need to engage with a user on a daily basis. Knowing the likes and dislikes of the user, the chatbot could keep the user stay engaged as well as better persuade the user to fulfill the required tasks, such as adhere to a particular care treatment. Not only will such a chatbot improve user engagement, but it will also improve care outcomes.

Make Responsible Chatbots

With great power comes great responsibility. Now that a chatbot can deeply understand its users, how can we ensure the responsibility of such a chatbot? For example, if a chatbot detects a shopper's impulsive buying behavior, how can we make sure that the chatbot does not indulge the shopper to buy more? Similarly, if a chatbot detects a gamer's addiction to violent games, how can we make sure that the chatbot does not encourage the gamer to become more addicted? In the near future, I will blog more about how to make responsible chatbots that can truly help people but not harm them.

A Step to Step Guide: How to Make a Multi-Purpose Chatbot with No Coding

2020-09-30T21:59:01Z

The Chatbot Tsunami has brought us a flood of chatbots to help automate various business functions, including customer service chatbots that automate customer Q&A, marketing research chatbots that automate customer interviews, and HR chatbots that automate job interviews. Although a chatbot is often made to serve one purpose, users expect the chatbot to perform multiple related functions while serving the main purpose.

Here are several examples.

Chatbot for Insurance Company Website

You want to make a chatbot for an insurance company website. Although this chatbot's main purpose is to provide site visitors with an insurance quote, it must also answer user questions during the process (see chat example as below).

Chatbot for Hospital Website

You want to make a chatbot for a hospital website. Although this chatbot's main purpose is to triage care situations, it must also answer visitor questions during the triage process (see chat example below).

A Chatbot for Bank Website

You want to make a chatbot for a bank website. Although this chatbot's main purpose is to help customers manage their accounts, it must also answer customer questions during the account management process (see chat example below).

A Chatbot for HR Recruitment Website

You want to make a chatbot for a job recruitment website. Although this chatbot's main purpose is to collect job applications, it must also answer applicant questions without derailing the application process (see the chat below).

Challenges of Supporting Multi-Purpose Chatbots

All the above examples show that the best chatbots are the ones that can serve multiple purposes and handle the interleaving of different tasks, such as answering a user's questions in the middle of performing another task. However, building such a chatbot is non-trivial due to two main challenges described below.

Tracking Conversation Context

The first challenge is to keep track of a conversation context (task status and flows) and maintain such a context. As shown in the first example above--the insurance chatbot example, before finishing gathering a user's information to make an insurance quote, the chatbot was interrupted by a user's question. Note that answering the user question also involves multi-turn interaction, which could also be interrupted by additional questions. In such cases, a capable chatbot should answer the user's question and then resume the original chat flow to finish the quoting task. If the chatbot cannot track and maintain a task context, the user would not be able to obtain the quote or must restart the whole process again. When that happens, we can imagine the kind of frustrations the user might endure.

In practice, any multi-turn task flow can be interrupted arbitrarily--at any step, any time, and any depth of a conversation (e.g., by another multi-turn tasks, which could be interrupted again). Keeping track of such a dynamic and complex context means remembering every conversation state, including task status and task flows, which is non-trivial. Most chatbot platforms leave chatbot designers or developers to manually tracking a conversation context, which is impractical or even infeasible given the number of the potential interruptions.

Maintaining Conversation Context

In addition to tracking a conversation context, a capable chatbot must maintain the context properly so it can resume a task flow after each interruption. To resume a context, a chatbot must inform users where they are (e.g., which task and where in the task) so the users are aware and can continue with the resumed task flow.

Using the chat examples shown above, each chatbot must repeat what it was saying (e.g., re-asking its question) when resuming from an interruption. This is similar to defining "reprompts" using Google Dialogflow. The challenge however is that if a chatbot designer must manually define "reprompts" for every possible interruption point in a conversation.

Again, most chatbot platforms leave the challenges of maintaining and resuming a conversation context to chatbot designers or developers, who must painstakingly recognize all potential interruptions and add "reprompts" to resume a conversation. This process is also impractical and infeasible as the number of tasks or intents grows.

3-Step Solution to Multi-Purpose Chatbots

To relieve chatbot designers or developers from manually tracking and maintaining a conversation context, Juji offers a solution that enables chatbot designers/developers to make a multi-purpose chatbots in three steps with no coding.

A Practical Chatbot Example

Here we use a practical example to illustrate the solution. Our example is to make a simple but practical chatbot for an insurance company website. As shown below, this chatbot will accomplish two main goals: (1) helping site visitors obtain a car insurance quote in 4 steps; and (2) answering certain visitor questions during a conversation.

Like shown in the first example above, the chatbot should support arbitrary interleaving of tasks and question answering. Before you read the solution below, I encourage you to first try to use a chatbot platform that you are most familiar with to implement this chatbot and see how long it will take you to accomplish both goals described above.

Step 1: Define a Main Chat Flow (Chatbot Initiated Tasks)

While it may take you days if not weeks to make such a chatbot using a normal chatbot platform, it will take just three key steps to build such a chatbot on Juji with no coding.

The first step is to define a chatbot's main task flow. This includes all the tasks your chatbot will initiate. In our example, there is only one task that the chatbot will initiate: helping a visitor obtain a car insurance quote. Defining such a task flow in Juji is rather straightforward, similar to using any of the online survey software like SurveyMonkey to enter survey questions. As displayed below, just enter each of the three questions.

Unlike a plain question in a survey form, in a Juji chatbot, a pre-built mini conversation is behind each entered question. For example, behind the question, "What's your age", a mini-conversation is used to automatically handle diverse user age input ("My age is 23" or "I'm 35 year old") and potential digressions (e.g., "Why do you need my age"). Because this pre-built mini conversation automatically handles diverse user intents and stores legit user answers, Juji relieves a chatbot designer from manually doing all the work. A chatbot designer can of course customize a pre-built mini conversation if needed.

Step 2: Define Q&A List (User Initiated Tasks)

After defining the main chat flow that includes the tasks that a chatbot will initiate, the second step is to define tasks that users will initiate during a conversation. This includes user questions and comments. On Juji, defining a chatbot Q&A list is also straightforward as one can simply upload a CSV file that contains Q&A pairs or directly enter Q&A pairs in a GUI table:

This blog details the step-to-step process of defining the Q&A list, including defining multi-turn Q&As. The example below shows a multi-turn Q&A that is defined to help a user subscribe to an email list.

Juji also supports the sharing of tasks between the main chat flow and multi-turn Q&A flow. For example, one can define a Q&A pair with a question like "Could you give me a car insurance quote". The answer of this question can be defined as a multi-turn interaction that reuses what's defined in the main flow (see below).

The difference between the main flow of providing an insurance quote and this multi-turn Q&A is that the multi-turn Q&A can be invoked by users any time during a conversation even after the main flow is completed. This is very useful. For example, initially when the chatbot starts the main flow, a user may not want to get a quote yet. However, the user might want to ask for a quote later during the conversation. This enables the user to do so whenever s/he is ready.

Step 3: Automatically Generate AI Chatbot

After both the main chat flow and Q&A list is defined, you can now proceed to step 3 to "Preview" your chatbot. This will automatically generate an AI chatbot. This chatbot will activate the main chat flow first and it can be interrupted anytime during the main chat flow to answer user questions matching the Q&A list. After handling an interruption, it always automatically resumes the interrupted flow (no matter it is the main flow or a multi-turn Q&A) to ensure task completion (e.g., giving an insurance quote or finishing a multi-turn Q&A). If a user question does not match with those on a Q&A list perfectly, the chatbot will make recommendations, which informs users what it can do.

In other words, this generated chatbot tracks and maintains a conversation context automatically, relieving chatbot designers or developers from worrying about how a chatbot should be wired to manage different conversation threads and maintain a conversation context. In addition, updating or extending such a chatbot is very easy. For example, we can update the chatbot main flow with different tasks or extend the Q&A list with more Q&A pairs without worrying about how they may interleave of each other during a conversation (e.g., no need to define specific "reprompts" for added tasks or Q&As). This is because a Juji chatbot uses algorithms to automatically detect and manage a conversation context, no matter how the main chat flow or the Q&A list changes.

A Step to Step Guide: How to Power Customer Service Chatbots with AI (No Coding Required)

2020-07-29T20:36:12Z

As indicated by this Forbes article, 76% of customers contact businesses to make inquiries and get support via text messaging. Now with the required social distancing during the COVID-19 pandemic, more and more consumers choose to interact with brands online.

Many businesses now receive hundreds or even thousands of customer inquiries daily. Let's assume that each customer service agent could handle about 30 calls per day, doubling the estimate in this report, it would need 10 human agents to answer about 300 incoming customer inquiries. On the other hand, short of manpower implies the prolonged customer waiting time and a Forrester study shows that 2 out of 3 customers don't want to wait for more than 2 minutes to get assistance.

To cope with the increasing volume of customer interactions, businesses started to look for solutions to automate customer interactions. One promising solution is to use a chatbot to augment human agents and help answer repetitive questions, such as "How to reset my password", "What is your refund policy" and "Where can I download your mobile apps". Even better, a chatbot can be trained to handle complex questions that require multi-turn interactions, such as "Can you find an online tutor for me?" and "Could you show me how to use your product?"

Challenges for Making Customer Service Chatbot

However, setting up a capable chatbot for customer services may be non-trivial, since a customer service chatbot must have a set of key capabilities to deliver satisfactory customer experiences. From our own experiences of helping organizations set up and manage customer service chatbots, we have discovered that a customer service chatbot often fails to fulfill its purpose because one of the following reasons.

Lack of Natural Language Processing (NLP)

If a chatbot does not have any Artificial Intelligence (AI) capabilities, it won't understand a user's natural language inquiries. A chatbot without AI or NLP is most likely to force users to traverse a tree-based menu by clicking on the displayed choices one by one before finding or not finding the answers they want. This approach never works well even if your business is super simple for two reasons. First, such a tree-based menu is often limited with choices and can hardly include every answer to customer questions. Second, customers may not know which path to take or which option to select.

For example, on an e-learning website, I want to find out which SEO course would be suitable for a beginner like me. It would be hard to insert such an answer in a tree-based menu. In another example, if I want to know whether I can get a free trial of a particular product, should I traverse the path to "Product" or "Price"?

It is worth mentioning an often used "NLP" technique supported by some chatbot platforms is keyword spotting, which retrieves potential answers based on certain keywords mentioned in a user's question. Unfortunately, keyword-based processing is often inadequate for a customer service chatbot because it cannot distinguish the semantic differences between customer questions containing the same keywords. Consider these two questions both containing the keyword "refund":

"Can I get a refund for the course I paid for"

"When will I receive my refund"

Users would not be happy if they ask the latter question but get the answer to the former question, or vice versa.

Here is a short Youtube video showing the difference between understanding the keywords vs. the semantics of user input.

Lack of Abilities to Handle Interruptions

As a customer, you may have experienced such a frustration with an incapable chatbot: you asked for a service, e.g., getting an insurance quote, and you had followed the chatbot instructions down for 5 steps already. But you had a question near the last step so you asked the question. Poof, all you had done so far was lost and you had to start over again. That is because such a chatbot does not know how to handle out-of-flow interruptions or has no concept of context. Once it is interrupted, the chatbot could not resume its original flow but restarts. As you can imagine, how frustrated a customer could be when encountering such a situation! Again, the same Forrester study shows that 63% customers abandon a company after just one bad experience.

Lack of Abilities to Help Human Agents

No chatbot is perfect and it cannot answer every customer question. It thus is very important for a chatbot platform to have the abilities to monitor a chatbot activities in real time, know when a chatbot fails, and notify human agents of chatbot failures so they can help address the gap.

Ideally, a chatbot platform should also help a human agent better handle the handed-over, unresolved issues, such as suggesting potential answers or knowing the emotional state of the users involved. Many chatbot platforms support chatbot-human handoff, but few can truly aid human agents in resolving the issues efficiently let alone understanding the users involved.

Difficult to Setup and Update AI Chatbots

Some chatbot platforms may support all the three capabilities mentioned above. However, it may take a team of engineers months to set up a customer service chatbot and weeks to update such a chatbot, e.g., updating the chatbot's knowledge base to improve its performance. Moreover, most platforms require AI expertise (e.g., defining intents and entities) or programming experience (e.g., manually wiring all different intents and re-prompting) to power a customer service chatbot with AI capabilities described above. With such a long turnaround time and difficulties in the chatbot upkeep, organizations may hesitate taking the steps to consider let alone implement a customer service chatbot.

Three Steps to Power Customer Service Chatbots with AI

To address the challenges in the adoption and upkeep of customer service chatbots as discussed above, Juji platform is designed to enable organizations to build and manage effective customer service chatbots rapidly. Specifically, the Juji Studio radically simplifies the teaching of AI and enables non-IT professionals to build and manage powerful AI chatbots for customer services in three steps, and no coding required.

Step 1. Prepare a free-text Q&A list

The main function of a customer service chatbot is to answer customer inquiries. As mentioned above, customer inquiries may be highly diverse, in both natural language expressions and topics. A tree-based menu can hardly satisfy customer needs. To deliver exceptional customer experience, a customer service chatbot should be able to understand customer free-text inquiries and respond accordingly.

To power a chatbot with free-text Q&A capabilities, Juji enables you to easily teach a chatbot in one of two ways. First, you can enter Q&A pairs directly in a table as shown below:

Once the entered Q&A pairs are submitted (using the "Submit" button), a chatbot is enabled for free Q&A.

With true NLP capabilities, Juji chatbots can recognize semantically similar user questions in diverse expressions. For example, with the above entered Q&A pair (e.g., "When will I receive my refund"), the chatbot is able to answer a user question expressed very differently ("How long will take my refund to arrive") as shown below.

Second, if you have a long list of Q&A pairs already, you can download our CSV template and put them into a CSV file as shown below. The format of the CSV file is also explained in our online documentation.

Design helpful HELP command

Once you enable your chatbot to support free-text Q&A, it is always a good idea to write a helpful HELP command that can guide your users' interaction with the chatbot.

Similar to a restaurant menu, a HELP command can help achieve two goals:

helps users get started especially if a user is new;
teaches users your chatbot's capabilities and sets up the right user expectations.

Here are 3 quick tips on writing a good HELP command for your chatbot:

Write a response that gives an overview of your chatbot
Paraphrase HELP responses to expose different but specific chatbot functions (see the screenshot below)
Make sure your chatbot does what it promises in its HELP command.

Satisfying the last point mentioned above is especially important to meet user expectations and deliver satisfactory user experience. In particular, you want to make sure the following conversation won't ever happen to your chatbot:

User: help
Chatbot: I can help you transfer money.
User: transfer money
Chatbot: I am sorry I don't understand you.

To prevent such user frustrations, you want to:

carefully word the HELP command responses to anticipate and minimize possible user input variations; and
handle potential user input variations that you can anticipate.

Convert website FAQs to Q&A entries

If you already have a webpage that includes FAQs about your business, you can also contact us and check if we could help you automatically convert these FAQs into the entries in a CSV file, saving your time and effort of preparing for the CSV file from scratch.

Step 2. Define multi-turn conversation flows to handle complex questions

Sometimes, a chatbot must engage with a user in a multi-turn conversation in order to answer the user's question. Consider the following example:

As shown above, the chatbot needs to collect basic information from the user first before answering the user's question.

To enable your chatbot to handle complex questions, Juji allows you to define a multi-turn conversation easily. In particular, you can define a mini chat flow that can elicit user information and use such information to decide how to best answer the user's question.

The screenshot below shows a chat flow defined to handle the user's request of seeing a doctor.

During such a multi-turn interaction flow, a user may interrupt the interaction flow with additional questions. For example, the chat flow below shows that the user asks a question about insurance before finishing up the thread on seeing a doctor.

As you can imagine, a user can interrupt an ongoing chat flow and activates a new one anywhere, anytime. In other words, a chat flow can arbitrarily nest one or more other chat flows. Fortunately, Juji automatically manages all chat flows and keep track of the conversation context so you don't need to manually track the conversation context or wire different chat flows together.

Step 3. Monitor and improve customer service chatbot periodically

There are always questions that a chatbot does not understand and cannot answer. A good AI chatbot should be able to handle unknown user questions as well as notify human agents where a chatbot fails. To better help human agents, Juji provides a Q&A dashboard that lists unanswered user questions as shown below.

Whenever Juji detects an unanswered user question during a chat session, it lists the question on the Q&A dashboard. Juji also automatically suggests a potentially matched answer (as highlighted in red in the answer column above) if an unanswered user question is similar enough to an existing question in the knowledge base (e.g., exceeding a similarity threshold).

To keep a customer service chatbot up-to-date and improve the chatbot over time, we encourage a human agent or chatbot owner to check the Q&A dashboard periodically (e.g., once a day) and perform two actions:

check the Q&A dashboard to verify the suggested answer or provide proper answer to each unanswered user question; and
submit the Q&A pairs to update the chatbot

Once the the Q&A pairs are submitted, the chatbot is updated immediately and can answer similar questions in any ongoing chats.

How to Evaluate Customer Service Chatbot Platforms

As described above, Juji supports an easy 3-step process for an organization to build and upkeep a customer service chatbot with powerful AI capabilities and no coding required. Since chatbot platforms are abundant, in case you wish to compare different chatbot platforms before deciding on which one to use to build your customer service chatbot, we suggest that you evaluate a platform from four aspects to maximize your chatbot ROI:

Customer Service Quality. This is to measure how well a chatbot can serve your customers. Can the chatbot understand user questions in diverse natural language expressions on a wide range of topics? Can the chatbot handle complex user questions that require a multi-turn conversation? Can the chatbot handle arbitrary user interruptions occurred during a chat while still ensuring task completion?

Ideally, you want your chatbot to automatically handle at least 40% customer inquiries from the very start and gradually improve to handle over +70% customer service requests over 1-2 months. This will present over 60% savings on your customer service cost and infinite amounts of customer happiness.

Implementation Speed. This is to measure how fast you can set up a quality chatbot with full AI capabilities as discussed above to deliver satisfactory customer services. Ideally,

for a small business with thousands of customers, the setup time should be within 1-2 days, not for weeks; and
for a medium business with hundreds of thousands even millions of customers, the setup time should not exceed 1-2 weeks, definitely not months.

The speed to setup often reflects the effort required to upkeep your chatbot and eventually the ROI of your chatbot. If it would take months to set up an initial customer service chatbot, it is most likely that it will take weeks to make any changes in your chatbot. Another associated criterion is to know who needs to be involved in the chatbot setup and upkeeps. If everything requires the involvement of your IT team, it is a signal to a very costly solution with low ROI.

If it is difficult to keep a chatbot up to date, customers will then eventually abandon the chatbot and the chatbot investment will go down the drain.

Service Robustness and Handoff. This is to measure how well a chatbot can handle exceptions -- unknown user inquiries. No chatbot is perfect. You want a customer service chatbot to be resilient even under uncertainties, which directly impacts customer satisfaction.

For example, what does a chatbot do when it encounters unknown user questions or comments? Does it recommend similar questions and give users a chance to find the answers or simply reply with "sorry, I don't understand your question"?

Another consideration is how easy to support chatbot-human handoff. Human agents should be aware of chatbot failures, ideally all time and in real time so they can help resolve the issues promptly. To deliver exceptional customer experience, human agents should have sufficient information for the handoff, ideally understanding emotional state and unique characteristics of the customers involved beyond the issues to best serve the customers.

Chatbot Learnability. This is to measure how fast and easy a chatbot can be improved. Adopting a chatbot is like raising a child, you'll have to update the knowledge of a chatbot periodically to improve its performance. Selecting a platform that can help human agents to update a chatbot easily and quickly can greatly improve the ROI of a customer service chatbot.

From Customer Service to Customer Advocacy

Another criterion you may want to consider is how easy you can extend a customer service chatbot to becoming a customer advocacy chatbot. In addition to answering customer inquiries, a chatbot could also elicit customer reviews or gather user feedback about products and services in the same chat (see screenshots below).

Juji allows you to do so easily without coding by adding a few conversation topics in the chatbot main flow. The added conversation can further improve the ROI of a customer service chatbot by automating more customer operations (e.g., product survey).

How to Improve Q&A Chatbots with Real-Time Monitoring, Notification, and Enhancements

2020-07-12T03:34:55Z

Since one of the most popular chatbot uses is to automate customer service, a chatbot's abilities to answer user questions is directly related to brand image and customer satisfaction. Previously, I talked about how to create a Q&A chatbot in a few minutes to answer users' free-text questions and handle complex, multi-turn Q&As, all without coding. Because no chatbot is perfect, I also mentioned how to teach a chatbot handle unknown user questions and further improve customer satisfaction.

But how do we know the performance of a Q&A chatbot? For example, we want to know the "goodness" of a chatbot by asking:

What are the questions the chatbot has answered? How many of them?
What are the questions the chatbot cannot answer? How many of them?
Among the unanswered questions, how frequently do users ask these questions?
How are these unanswered questions related to each other?

Obviously, answers to the above questions can provide chatbot owners or admins valuable insights to improve a chatbot, e.g., improving a chatbot to address the most frequently asked, unanswered questions. Unfortunately, few chatbot platforms provide such insights, which makes a chatbot owner/admin unaware of a Q&A chatbot's performance, let alone helps improving its Q&A capabilities.

To help chatbot owners improve a Q&A chatbot easily and quickly, the Juji platform provides built-in Q&A monitoring facility to inform chatbot owners of their chatbot Q&A performance in real time.

Monitor Overall Chatbot FAQ Performance

Besides monitoring unanswered questions, one can also monitor a chatbot's overall FAQ performance on the Reports page.

As shown above, it lists the number of questions users asked, and the number of questions answered by the chatbot. It also calculates the answering rate. This metric informs a chatbot owner or admin how well a chatbot has performed in answering questions. A low answering rate may be a strong indicator that a chatbot's FAQ abilities need to be updated by checking the Q&A dashboard and addressing the unanswered questions as much as possible.

Using the data, we can also calculate the average number of questions asked per user by the following:

AVG NUM QUESTIONS PER USER = NUM of QUESTIONS/NUM of USERS

Monitor Q&A Chatbot Performance: Monitoring Unanswered User Questions

Additionally, we have built several Q&A chatbot monitoring tools to answer questions 2–4 listed above. Here I focus on how Juji helps chatbot owners obtain answers to these questions and how such insights can help improve a chatbot's Q&A capability quickly.

Ideally, your FAQ list should be as complete as possible. However, it is difficult to anticipate all the questions users might ask. Thus, it is useful to know what user questions are not answered by a chatbot. Juji's Q&A Board does just that. As shown below, the Q&A board provides you with real-time notifications of unanswered user questions as users are interacting with your chatbot.

As shown in the screenshot above, the number shown in red next to the label "Q&A BOARD" indicates the number of unanswered user questions the chatbot has encountered. The number gets updated in real-time. So a chatbot owner can easily know how many unanswered questions are there. In the example above, there are 5 unanswered user questions. Moreover, all the unanswered questions are listed inside a Q&A table where a chatbot owner or admin can answer them right away.

Alternatively, you can download all Q&As including the unanswered questions in a CSV file. The CSV file typically lists all the unanswered user questions and their stats at the end. Juji also automatically analyzes all unanswered questions and arranges them in different groups by their semantic similarity.

As shown in a sample CSV file below, there are five unanswered questions (rows 38-44), arranged into three groups: rows 38-39, rows 41-42, and row 44. Next to each unanswered question, there is also a number indicating how many times (frequency) that question was asked. In the example below, the question "How do I log in" (row 38) was asked once, while another similar question "How do I sign in" was asked twice (row 39).

Such information can then be used to easily prioritize the handling of unanswered questions especially if there are a number of unanswered questions.

Update Chatbot FAQs

To address unanswered user questions, there are two ways to improve a chatbot by updating its FAQ knowledge base. You can update them directly on the Q&A Board. Alternatively, you can update the CSV file, then upload it to the chatbot.

Update FAQs on the Q&A Board

If there are just a few unanswered questions, the quickest way is to add the corresponding answers directly on the Q&A board. To do so, just add an answer in the corresponding text box (labeled as "Chatbot Responses") and submit it using the "Submit" button.

Unless this is a multi-turn answer, the submission will take effect immediately. This is very handy, because it improves a chatbot Q&As immediately without the need to restart it.

Answer an unanswered user question

Assume that an unanswered user question pops up on the Q&A Board. As a chatbot owner, you can answer it and submit the answer right away. ALL the users including the one who asked the question originally can now ask the question and get a correct answer without restarting their chat. This nondisruptive nature of the updates has proven to be very valuable to many of our clients that have a large number of users.

As shown above in a screenshot, the Q&A Board also provides several useful tools to help a chatbot owner customize answers:

The "+" sign in a text box (either the "User Questions" or "Chatbot Responses" column) allows you to enter alternative question/answer expressions, while the downward arrow button allows you to view and select these varied expressions. Question variations help a chatbot better recognize user questions. On the other hand, answer variations give a chatbot a set of answers to choose from, so its response can appear more natural to users.
The "→A" button on the top right corner of the "Chatbot Responses" text box lets you search for an answer of a matched question, saving your time to enter an answer from scratch. This is especially helpful if there is an extensive existing Q&A list.
The "[+]" button on the bottom right of the "Chatbot Responses" text box lets you open up an editor to define or edit a multi-turn answer. My previous blog post explained this powerful tool.

Add a new pair of Q&A

Moreover, the Q&A Board allows you to directly add new Q&A pairs. To add a new pair of Q&A, simply click on the "+" sign at the top of the table. You can then edit the question and its answer as described above.

Update an existing Q&A pair

Similarly, you can select an existing Q&A pair to edit by clicking on the magnifying glass icon next to the "+" sign. As shown below, it will open up a pop-up window where you can search for an existing Q&A. In the pop-up window, you can see all Q&As including their question and answer variations. Once you select a Q&A, it will show up in a row in the Q&A table for you to update.

Update and upload your FAQ CSV file

Although the Q&A Board makes it very easy to update your FAQs, updating the FAQs inside the CSV file may be more convenient if there is a large number of unanswered questions or if there are multiple people involved in updating the FAQs.

You can simply upload your CSV after all the editing is done. In case you are not already familiar with how to update the FAQs in a CSV file, you can check out my previous blog on this.

Answering user questions is an important skill for almost all chatbots. That's why Juji Q&A dashboard exists to help chatbot designers and owners monitor their chatbots in real time and improve their chatbots quickly.

Creating a chatbot and publishing it is not the end but often the beginning of a journey. Just like raising a kid, who needs to be taught and nurtured as s/he grows, a chatbot designer or owner will need to keep a chatbot up to date and constantly teach it new knowledge and skills. Over time, a chatbot can become better and better at its tasks.

If you would like to learn more about what kind of smarts Juji chatbots have, this is a great article to start. Or sign up free to have some fun of creating your own smart chatbots.

Comparing Clojure Diff Libraries

2020-07-11T23:43:37Z

In my Clojure/north 2020 talk on "diffing-based software architecture patterns", I mentioned that Juji is using Editscript, a library I developed, to diff Clojure data structures. During the Q&A session of the talk, someone brought up another Clojure diff library, called deep-diff2, which I was unaware of. Then on Youtube, a comment asking the difference between Editscript and deep-diff2 appeared again. This prompted me to do an investigation on Clojure data diff libraries. Given how the Clojure community places such an emphasis on data oriented programming, a comparison of data diff alternatives appears to be of interest.

My Google search with "Clojure diff" brought up these options: clojure.data/diff, differ, and aforementioned deep-diff2. Curiously, Editscript does not even show up on Google, despite the fact that it has the most number of github stars among all the options. Anyway, I would like to do a comparison among these options. Before going into the details of the differences, how about doing a benchmark first?

Benchmark

All these library implements a diff function, so we can measure how long it takes for them to diff the same pair of data structures. We will also compare the sizes of the resulting diffs.

Data Set

I happen to have a data set copied from a JSON diff benchmark, which seems to include data models of a Javascript drawing program. I simply converted the JSON files into Clojure EDN format. There are four files, data1, data2, data3, and data4, with each being a variation of another. After serialized into bytes with nippy, the sizes of the data are 1004, 1004, 1016 and 555 bytes, respectively.

The shape of the data has a bit of challenge for diff algorithms. The top level is a vector of nested maps. It is necessary to maintain the order of the vector elements, at the same time, the algorithms need to dig into each pair of nested maps to find the differences. However, the nesting is not deep.

Test Environment

Since these alternatives are all Clojure libraries, I created a project to simply pull the latest versions of them from clojars and let them loose on the data set. The code is here. The timing benchmark uses criterium quick-bench function.

The test ran in a Clojure REPL on a Ubuntu Linux 5.4.0-58-generic server with Intel(R) Core(TM) i7-6850K CPU @ 3.60GHz, and JVM is OpenJDK 64-Bit Server VM AdoptOpenJDK (build 15.0.1+9, mixed mode, sharing)

Results

As you can see, the time for the libraries to run diff algorithm on the dataset varies greatly. The same is true for the resulting diff sizes.

To choose a library, we need to look at each library to see if it fits one's use cases. In addition to the performance data, we also need to look at the output format of each library.

Analyses

Let's look at each option individually first, then do a summary.

clojure.data/diff

This is a built-in function of Clojure. The doc string says:

Recursively compares a and b, returning a tuple of [things-only-in-a things-only-in-b things-in-both]

Obviously, this simple walking-through of two data structures is not meant to uncover the minimal differences between them. The resulting diff will always be larger than the original size of the data. The diff size chart above shows just that. Here's what the result looks like:

(pp/pprint (clj/diff data1 data2))
;;==>
[[nil nil {:fill "#ffff00"}]
 [nil nil {:fill "#0000ff"}]
 [{:y 27,
   :r 0,
   :color "#000000",
   :fill "#CCCCCC",
   :width 100,
   :type "rect",
   :cap "round",
   :borderWidth 1,
   :style "Solid",
   :x 50,
   :height 100}
  {:y 30,
   :family "sans-serif",
   :color "#0000FF",
   :fill {:r 256, :g 0, :b 0, :a 0.5},
   :width 10,
   :type "textBlock",
   :cap "round",
   :borderWidth 1,
   :size "24px",
   :style "Solid",
   :pad 3,
   :weight "bold",
   :x 20,
   :height 25.200000000000003,
   :text "DojoX Drawing Rocks"}
  {:rx 150,
   :color "#0000FF",
   :type "ellipse",
   :cap "round",
   :borderWidth 1,
   :style "Solid",
   :cx 150,
   :cy 185,
   :ry 100}
  {:color "#000000",
   :fill "#CCCCCC",
   :y1 20,
   :type "arrow",
   :cap "round",
   :borderWidth 3,
   :style "Solid",
   :label "My Arrow",
   :x1 40,
   :y2 120.00000000000003,
   :x2 -133.20508075688772}
  {:y 26,
   :family "sans-serif",
   :color "#000000",
   :fill "#CCCCCC",
   :width 200,
   :type "text",
   :cap "round",
   :borderWidth 1,
   :size "18px",
   :style "Solid",
   :pad 3,
   :weight "normal",
   :x 30,
   :height 25.200000000000003,
   :text "This is just text"}
  {:color "#000000",
   :style "Solid",
   :cap "round",
   :fill "#CCCCCC",
   :borderWidth 1,
   :points
   [{:x 70, :y 20}
    {:x 65, :y 15}
    {:x 75, :y 15}
    {:t "Z", :x 70, :y 20}
    {:t "M", :x 70, :y 40}
    {:x 68, :y 12}
    {:x 72, :y 12}],
   :type "path"}]]

In term of speed, this function is not the best. When diff is small, it is in the middle of the pack. When the diff is large, it can be the second slowest, only beating the optimizing A* algorithm of Editscript, but that algorithm is doing much more than simply walking two trees.

This function does produce a lot of information about the changes. There may be cases when these are useful. For example, you can ignore "things-in-both". "things-only-in-b" tells you the new things added, and "things-only-in-a" tells you what are deleted. However, if both are not empty, it would be hard to figure out what exactly happened.

There is no corresponding patch function for diff, so you cannot really use this to preserve and restore data.

lambdaisland/deep-diff2

This library seems to enjoy a lot of attention. Here's the tag line:

Recursively compare Clojure or ClojureScript data structures, and produce a colorized diff of the result.

So it seems to gear towards visualizing the data diff for human consumption. Here's what the results look like:

(deep/pretty-print (deep/diff data1 data2))
;; ==>
[{:borderWidth 1,
  :cap "round",
  :color "#000000",
  :fill "#CCCCCC",
  :height 100,
  :r 0,
  :style "Solid",
  :type "rect",
  :width 100,
  :x 50,
  :y 27}
 {:borderWidth 1,
  :cap "round",
  :color "#0000FF",
  :family "sans-serif",
  :fill {:a 0.5, :b 0, :g 0, :r 256},
  :height 25.200000000000003,
  :pad 3,
  :size "24px",
  :style "Solid",
  :text "DojoX Drawing Rocks",
  :type "textBlock",
  :weight[0m "bold",
  :width 10,
  :x 20,
  :y 30}
 {:borderWidth 1,
  :cap "round",
  :color "#0000FF",
  :cx 150,
  :cy 185,
  :fill -"#ffff00" +"#0000ff",
  :rx 150,
  :ry 100,
  :style "Solid",
  :type "ellipse"}
 {:borderWidth 3,
  :cap "round",
  :color "#000000",
  :fill "#CCCCCC",
  :label "My Arrow",
  :style "Solid",
  :type "arrow",
  :x1 40,
  :x2 -133.20508075688772,
  :y1 20,
  :y2 120.00000000000003}
 {:borderWidth [36m1,
  :cap "round",
  :color "#000000",
  :family "sans-serif",
  :fill "#CCCCCC",
  :height 25.200000000000003,
  :pad 3,
  :size "18px",
  :style "Solid",
  :text "This is just text",
  :type "text",
  :weight "normal",
  :width 200,
  :x 30,
  :y 26}
 {:borderWidth 1,
  :cap "round",
  :color "#000000",
  :fill "#CCCCCC",
  :points [{:x 70, :y 20}
           {:x 65, :y 15}
           {:x 75, :y 15}
           {:t "Z", :x 70, :y 20}
           {:t "M", :x 70, :y 40}
           {:x 68, :y 12}
           {:x 72, :y 12}],
  :style "Solid",
  :type

Sorry that this page does not do justice for the colorized output. But the thing to notice is the change :fill -"#ffff00" +"#0000ff".

So basically this library displays the data, then highlights the changes in color. Of course, the result size will be larger than the original data. When changes are significant, the size could be be more than doubled, as shown in the chart.

Speed-wise, this library is also only consistently faster than the optimizing A* algorithm. It sometimes beats clojure.data/diff when the diffs are large. This is remarkable, because what it does is a lot more than simply walking the trees.

Its credits section cites:

This library builds upon clj-diff, which implements a diffing algorithm for sequences, and clj-arrangements, which makes disparate types sortable."

I am familiar with the algorithm [1] used in clj-diff. It is an O(np) algorithm for diffing strings, where p is the number of deletes. The way it works is by maintaining a moving window of approximate size p along the diagonal of the editing matrix (think data a as the row, data b as the column), so it avoids searching the whole matrix. This algorithm is also implemented (with slightly better performance) in both versions of Editscript's two algorithms, to handle the special cases when we know simple sequences are being compared.

Clojure data structures are trees, not simple sequences of elementary values. The above algorithm assumes that each edit operation has the same cost, which is false for tree editing. Adding a large sub-tree costs a lot more than adding a single value by putting a lot more things in the resulting diff, for example. Another problem with that algorithm, is that it does not have replacement operator, having only add and delete operators. In any case, if optimal diff is desired, a proper tree diff algorithm is necessary.

However, general tree diff is expensive. The optimal time complexity is proved to be O(n^3) [2]. Fortunately, Clojure immutable data structure diff does not need or want general tree diff, where everything can move around. We actually want to preserve our beloved immutable data structures. This is how Editscript's A* algorithm can achieve optimality with less than O(n^2) time complexity: our definition of optimality disallows certain operations, such as splitting or merging nodes.

Editscript (A* algorithm)

This is the default algorithm of Editscript library. The reason I made this choice is because the optimal diff is likely the true diff. The library is intended as a part of the data transport for communicating software components (see my talk), where the content of diffs may control application logic. As such, they'd better be reflecting the true changes.

As mentioned, our structure preserving requirement makes our diff problem simpler than a general tree diff problem. We can now compare things layer by layer. Consequently, the time complexity goes down to O(n^2). On top of that, we implement A* search algorithm to obtain some further speed up. Basically, our A* algorithm uses a simple heuristic to avoid searching the whole edit matrix on each layer. This simple heuristic is actually inspired by the sequence diff algorithm mentioned above, i.e. it's about the number of deletes. However, in order to ensure optimality, our heuristic is currently rather conservative, so we may have left some potential savings on the table.

For now, I am fine with the performance. The benchmark shows that although it is the slowest, as expected, but in the big scheme of things, the cost is only a few milliseconds, something a lot of applications can bear.

Here is what the diff produced looks like:

(editscript/diff data1 data2)
;;==>
[[[2 :fill] :r "#0000ff"]]

That's it?

Yeah, that's it. That's the true change. The user apparently changed the fill color of the third shape.

Editscript (Quick algorithm)

This simple change happens to be also detected by the quick algorithm.

(editscript/diff data1 data2 {:algo :quick})
;;==>
[[[2 :fill] :r "#0000ff"]]

The benchmark shows that it is the fastest algorithm most of the time, only occasionally slightly behind the library that we will discuss in the next section.

This algorithm mostly does a one-pass walking-through of two trees, and notes any differences found. As mentioned, when we see a sequence, however, we activate the sequence diff algorithm mentioned above to gain a little bit of optimality. So at least in this case, this quick algorithm does not produce catastrophically large diffs. The kind of mistakes this kind of naive one-pass walking-through algorithms make, is that it would often delete a whole subtree, then add its slight variation back. Such edits are unlikely to be the true changes.

Editscript produces something does look like a script. The corresponding patch function takes the script and runs it to restore the data. The patching process takes almost no time, so the benchmark is omitted here.

The only other library that does patch, is differ.

Differ

By the numbers of the benchmark, this library looks amazing. It is almost as fast as the quick algorithm of Editscript, and its diff size is almost the same as the A* algorithm of Editscript! So we can have our cake and eat it too?

Sadly, there's no miracle in algorithms. It happened to be lucked out in this particular dataset, where the forbidden changes of this library did not happen:

Sequence items cannot change places
Set items cannot change themselves

So if one throws arbitrary data at it, differ will fall. The benchmark file contains code to demonstrate it failing all the property based tests, where the property checked is (= b (patch a (diff a b))). So unless you know your data do not involve the above changes, you probably want to use something else.

Summary

The Clojure community seems to care about data diff, so we have a few options. Based on what I have seen, if you are in the market to use a diff library, here are my recommendations:

If you want to look at the diffs and do not care if the diffs are always the smallest possible, deep-diff2 is great.
If you need to use the content of diffs in application logic, or you want to store the smallest possible diffs, and the necessary computing time is acceptable for you, use the default A* algorithm of Editscript.
If your data changes are too frequent or data is too big to accept the time cost of computing optimal diffs, and you do not care if the diffs may be sometimes not the smallest possible, the quick algorithm of Editscript may work for you.
I do not recommend clojure.data/diff, nor differ.

Regardless, test the libraries on your own data sets first.

[1] Wu, S. et al., 1990, An O(NP) Sequence Comparison Algorithm, Information Processing Letters, 35:6, p317-23.

[2] Demaine, E. et al. An Optimal Decomposition Algorithm for Tree Edit Distance. ACM Transactions on Algorithms, 6(1), 2009.

Where is the Artificial Intelligence (AI)? Detecting the Invisible AI that Makes a Chatbot Smart

2020-07-06T18:52:02Z

A while back, I wrote a blog on "Chatbot Tsunami", mentioning that our world is now inundated with chatbots, including both "good ones" that can help us fulfill our requests and "bad ones" that can hardly understand users or achieve anything.

Since Juji aims at powering "good" chatbots with the smarts needed to help users, I had assumed two things:

Everyone would know what these smarts are; and
Everyone could tell whether a chatbot has such or any smarts.

Until recently, I realized that my two assumptions were wrong.

A few weeks ago, I was demoing our conversational AI (AI chatbot) platform in a customer meeting. In this meeting, I showed how to build a chatbot as a personal healthcare assistant that can check in on a user daily. As shown in Figure 1, the chatbot (Clara) greets the user (Jane) and asks how Jane is feeling. Jane responds to Clara's greeting in natural language and the conversation continues.

Figure 2 shows another conversation between Clara, the chatbot, and a different user (Joe). Clara starts with the same greeting. However, Joe does not respond to Clara's greeting. Instead he asked Clara to help find a doctor. To help Joe, Clara starts to gather some basic information from Joe. Before Clara finishes gathering the info, she was interrupted again by Joe because Joe has a question about his insurance.

It took me just 3 minutes to create the chatbot Clara. I then spent a few minutes to show Clara in action (Figures 1-2).

Because everything worked flawlessly, I was thinking to myself they ought to be impressed by the smarts of our chatbot. However, there was a dead silence after my demo. I then asked "Do you have any questions about the demo?". One person then asked: "I got it. But where is the AI in this?"

Coincidently (or not so), in a separate meeting after this demo, another potential customer told me that "I tried Juji. While it is very easy to create a chatbot, I'm not sure where the AI is."

I was certainly taken aback by the question and comment. In addition, I felt quite disappointed at myself that I failed to highlight the hallmark of a Juji chatbot -- the chatbot smarts (AI).

The question and the comment then reminded me of some earlier comments I received. Some folks who read our documentation on the topic of "why AI chatbots" told me "by reading your documentation, it seems I don't need an AI chatbot." I didn't think too much then, and simply thought that they just need a "dumb bot" for simple tasks. Why bother to build a smart one?

Now I realized that all these comments are in fact related -- not everyone is aware of the smarts needed by a chatbot (why AI), let alone telling whether a chatbot has any smarts (AI) in it.

I feel compelled to use this blog to highlight the types of smarts (AI) that every chatbot should have and why. At the end of this blog, I will also share some quick tips on how to tell if a chatbot has any of such smarts (AI).

What is Artificial Intelligence (Definition of AI)

Before I outline the type of AI a chatbot should have, I briefly mention my definition of AI because it is directly related to the definition of AI chabot.

Although the definition of AI on wikipedia can provide more comprehensive information, my definition is pragmatic and short: AI is any human skills that we intend to teach a machine to have.

Four Smarts (AI) Every Good Chatbot Should Have

Since the purpose of a chatbot is to automate human interactions and help a user as if the user is interacting with a person, a good chatbot should have some essential human communication skills to carry on an effective conversation.

Let's imagine that a chatbot would behave like a person in a conversation. As outlined in the diagram above, such a chatbot should have four core human skills (smarts):

Natural Language Understanding (NLU). A true conversation is the exchanges of natural language expressions. This smart refers to a chatbot's abilities to interpret a user's natural language expressions and understand the user's intent(s).
Dialog Management. Once a chatbot understands the intent of a user, it needs to decide what to do -- how to best respond to the user and move the conversation forward***.*** For example, it may decide to answer a user's question in the middle of the flow (Figure 2) or to ignore the question (Figure 3). Because a user often does not follow a pre-defined chat flow, this smart often refers to a chatbot's abilities to manage user interruptions or digressions, keep track of a conversation context, and ensure the completion of a dialog (task).
Natural Language Generation (NLG). After a chatbot decides what to do, the chatbot then composes its response to the user. This smart refers to a chatbot's abilities to compose natural language responses in context. For example, a chatbot may compose different expressions to communicate the same content to different users.
Reading between the Lines to Understand Users Deeply (Personal Insights Inference). A good human conversationalist not only can grasp the intent of a conversation partner, but can also read between the lines to infer the partner's true needs and wants, as well as the partner's characteristics (e.g., cautious vs. easy going). If a chatbot can obtain such insights about its users, it can then personalize a conversation with each user. For example, a chatbot can recommend different books to different users based on their conversations.

The nature of a conversation is to exchange information and ideas expressed in natural language. Thus, understanding natural language expressions (NLU) is a basic smart that a chatbot should have. Although most chatbots have very limited NLU smart (Figure 4), there is much progress made on NLU especially with the recent advances in machine learning.

For example, Juji chatbots are able to understand highly diverse and complex natural language expressions using the hybrid pattern-based and data-driven machine learning methods. In contrast to the two chatbots shown in Figure 4, Figure 5 shows a Juji chatbot that is able to grasp a user's input correctly.

Similar to NLU, natural language generation (NLG) is another must-have chatbot smart because a chatbot must produce responses in a chat. NLG has also made great advances as summarized by Professor Kathy McKeown in her recent ACL 2020 keynote.

On the other hand, another two smarts dialog management and reading between the lines are less understood because fewer people are aware of their importance and there are also greater challenges for chatbots to gain these two smarts.

Next I will describe why it is important to power a chatbot with these two smarts.

Managing a conversation flow

Assume that a chatbot can understand a user's input perfectly. In such a case, the chatbot still needs to decide how to best respond to the user's input and move a conversation forward gracefully. Just like a person, a chatbot can carry on a conversation in many different directions depending on a user's input and the goals of a conversation. For example, if the goal of a conversation is to help users no matter what, a chatbot would act so. Figure 2 shows that Clara the chatbot switches the context twice to help user Joe and then picks up the flow to move a conversation forward. However, if the goal of a conversation is to elicit information from a user, a chatbot may ask the user to hold on his/her questions until the interview is finished (Figure 3).

Since it is very difficult to anticipate how a user would behave in a conversation, it is not possible to plan a chatbot's every move in advance. Moreover, many factors including human spontaneity will most likely prompt a user to digress from a planned conversation flow.

Figure 2 shows Joe's two digressions from a flow. Figure 5 shows another user digression: instead of replying to the chatbot with a name, the user asked the chatbot its name. Figure 6 shows yet another example where a pizza-ordering chatbot gets interrupted by the user's question about a coupon.

In all the examples above (Figure 2, Figures 5-6), it is important for a chatbot to handle user interruptions properly while moving the conversation forward. In all these examples, there is AI working behind the scene to drive the chatbots to manage the chat flow changes. In fact, a chatbot needs to make a decision at every step, e.g., deciding to answer a question or continue the current flow. Humans are super smart and can make such decisions in a split second, while a chatbot must be taught with such smart.

What would happen if a chatbot does not have such smart? The chatbot shown on the left of Figure 4 is not able to handle a user's interruption ("what's yours"). This inability also hinders the task completion (i.e., in this case, the chatbot failed to get the user's name).

No matter how simple a task is, a chatbot often needs such smart to ensure task completion. Let's say we want a simple chatbot to gather product feedback from our users. If we care about the quality of data gathered, we would be concerned with the following chat:

Chatbot: How did you like our product?

Participant: I don't know.

Instead of taking the participant's input, a good (AI) chatbot would have the smart to recognize the inadequate input and respond by encouraging the participant to give more meaningful input. Along the same line, what if the participant responded with one of the following expressions:

What aspects do you like to hear?
I haven't used the product long enough to have an opinion
Why do you want to know?

How should the chatbot manage the conversation?

As you can imagine, such "exceptional answers" could be many since user behavior is highly diverse and unpredictable, which can drive a conversation into numerous directions. To balance the delivery of a great user experience and fulfillment of conversation goals, a chatbot must have the smart to manage a conversation flow properly. On the other hand, a chatbot without such smart will have a hard time to complete any tasks, let alone satisfy its users.

Reading between the lines

Although many chatbot or conversational AI platforms mention that their chatbots can personalize a user experience, few of them actually do so. For example, if a pizza-ordering chatbot asks a user to make a series of choices and then orders a pizza based on the choices made, the conversation is not personalized because if another user who would make the same set of choices, the user will get the same pizza.

In contrast, if a user asks a chatbot to recommend a pizza to order, the chatbot makes a recommendation based on a user's likes and needs, that is a personalized experience. This is because if another user makes the exactly same request, she may get a totally different pizza based on her likes and needs.

Numerous studies in Psycholinguistics including our own (Section 4) have shown that a person's text communications reveal one's characteristics, such as motivations and preferences. A smart chatbot should be able to infer a user's characteristics from the chat, and then use the inferred insights to personalize each conversation.

For example, a chatbot can interact with a gamer to infer his gaming preferences, a social gamer vs. an action-oriented gamer, and recommend suitable games. Similarly, a chatbot can chat with a reader to infer her reading preferences, cerebral vs. artistic, and then recommend suitable books. Note that such recommendations are based on user characteristics, different from traditional behavior-based recommendation (e.g., collaborative filtering as seen on Amazon), which often requires large amounts of user behavior data. In contrast, characteristics-based recommendations can be made even if a user has no demonstrated behavior. Assume that a user is looking for a game to play but has never bought a game before. Through a brief conversation, a chatbot infers that the user is very artistic and highly curious. It could then recommend games that feature beautiful game worlds with intriguing plots.

In short, this smart enables a chatbot to know a user deeply and become truly personal to each user. This short Youtube video shows how to personalize chat messages based on a user's characteristics.

How to Tell if a Chatbot has AI

Based on the four must-have chatbot smarts described above, it is easy to detect if a chatbot has any AI. We can do so by interacting with a chatbot and answering four questions.

Can the chatbot understand your natural language expressions?

No matter whether you are evaluating a particular chatbot or a chatbot platform, I strongly encourage you to chat with the chatbot or a chatbot made on that platform (see more on platform evaluation).

Test on varied expressions. To detect if a chatbot can understand natural language expressions, try chat with it using different expressions and see how it behaves.

For example, if a chatbot asks you "What's your name?", try to answer it in different ways:

* my name is Michelle

* I go by Michelle

* Michelle is the name

* My name is not Mary but Michelle

This is to test its abilities to understand natural language expressions because your chatbot users will sure come up with many different expressions.

Ask a confirmation question. If a chatbot does not confirm its interpretation results, you also should test it by asking a confirmation, e.g., "What's my name" to confirm if the chatbot gets your name correctly.

Embed keywords into expressions with opposite meanings. To test whether a chatbot does only keyword matches instead of semantic interpretation, you can also try to input expressions containing the same keywords but with completely different meanings to see if you'd still get the same response. For example, if a chatbot asks "How healthy are you?" try "I'm not healthy" and "I'm very healthy" to see what chatbot responses you'd receive.

After several tries, you'll get a good idea as how smart a chatbot is in terms of understanding natural language expressions.

Can the chatbot manage a conversation?

Assuming that a chatbot can understand your natural language expressions, the next step is to check whether it can manage a conversation properly. You can test this smart by asking a chatbot questions at different points of a conversation and observe how a chatbot handles your interruptions:

Does the chatbot ignore or answer your questions?
Does the chatbot resume the conversation flow after answering your question?
Does the chatbot keep track of the conversation context (e.g., asking the chatbot "where are we")

The examples shown in Figure 2, Figure 5, and Figure 6 are good test cases to detect a chatbot's smart in dialog management. For the best chabots, you could even test how a chatbot handles multi-level interruptions as shown in Figure 2.

Can the chatbot be personal?

We suggest two ways to test whether a chatbot has the smart to personalize a conversation. One is to directly ask the chatbot "How much do you know about me?" and see how it responds (Figure 7). Second, if a chatbot makes recommendations or give advices, ask for a recommendation or advice. Then ask a friend of yours to do the same and see if you get the same results.

In short, I have described the four types of smarts (AI) every chatbot should have and how to detect if a chatbot has such smarts. Since it is non-trivial to power a chatbot with all the smarts, I will share how to do so in my next blog.

In the meantime, if you wish to test out the smarts of Juji chatbots, chat with them or sign up to create your own and experience such smarts.

How to Design Chatbots for Better Customer Service: Handling Unknown User Questions

2020-06-15T01:27:17Z

Customer Service Chatbots

One of the most intuitive uses of chatbots is perhaps to automate customer service, where a chatbot automatically answers user questions, 24x7. In fact, in almost every chatbot application, a chatbot's ability to answer user questions, especially free-text questions, is always desirable.

As shown below, a travel chatbot is used in a marketing campaign to introduce new travel products on Facebook Messenger. Customers who are interested in may have questions about the products, such as "What packages do you offer" or "Do you any promotions" .

If a chatbot can understand such user questions and respond to them immediately, it will turn a traditional, static marketing campaign into an instant live conversation, helping convert the visitors into customers, perhaps immediately. The screenshot below shows that the travel chatbot understands a user's free-text question and responds to it immediately.

Since 66% customers leave after waiting for 2 minutes to get an answer, a chatbot's always-on, immediate responses could really give a business the advantage of serving its customers. On the opposite, if a chatbot does not allow users to ask any questions, it will lose the precious opportunity to engage with the users "on the spot".

Three Chatbot Challenges for Customer Service

However, no matter how smart your chatbot is, there will always be user questions that it does not know how to answer. In general, there are three reasons why a chatbot cannot answer a user question:

the chatbot designer has not provided the chatbot with sufficient information to answer the question;
the user has not given enough information for the chatbot to answer his/her question;
the chatbot fails to understand the user question.

How to Design a Chatbot for Better Customer Service

Based on the three reasons listed above, a chatbot designer can use different methods to "fix" a chatbot and improve its abilities to handle user free-text questions.

Improve a chatbot with more knowledge

If it is the first reason, a chatbot designer wants to be notified whenever a user question is unanswered so that s/he can teach the chatbot to address the unanswered questions. This way a chatbot will become smarter and smarter as it gains more knowledge. In fact, this is the topic of my next blog on how to monitor unanswered user questions so a chatbot can learn and be improved over time. if you are interested in this topic, follow Juji's Linkedin page, which will feature each of our blogs.

Improve a chatbot with multi-turn replies

If it's the second case, a chatbot designer could ask users to supply missing information by adding a multi-turn reply to the question. As shown below, the travel chatbot in our example asks the user to provide additional information before it gives out an answer.

Improve a chatbot to handle unknown user questions

So how about the third case? What happens when your chatbot doesn't understand a user's question?

One approach is that a chatbot simply acknowledges that it does not understand the question. If you have used any of the smart voice devices, such as Amazon Alexa or Google Home, you might have already been used to hearing their apologetic responses when they don't understand your commands. Similarly, a chatbot could respond with something like "Sorry, I don't know how to respond to your question."

Question recommendation

While this "I don't understand your question" approach is straightforward, it does not really help a user get his/her question answered. Alternatively, another approach is to power a chatbot with a question recommendation engine, which will automatically suggest a set of similar questions to a user if a chatbot doesn't understand his/her question or does not have enough confidence to interpret such a question.

Let me use the following use cases to explain how such a mechanism works, more importantly, how it helps improve customer experience and better customer service.

Use Case

My friend Sarah hosts an educational website, which includes many fun, online educational games. She created a chatbot on Juji as her website assistant, who can guide her site visitors to navigate the site as well as answer their questions about the company and the games. Following our chatbot design tips, Sarah added a list of Q&As, to handle visitor questions like "Where do I log in" (see below).

Since an AI chatbot uses machine learning to interpret a user's free-text questions, it could recognize questions with similar meanings (semantics) even if the questions are expressed differently. For example, an AI chatbot can recognize that "how old are you" and "what's your age" mean the similar thing, although their expressions are very different. In other words, if an AI chatbot knows how to answer "what's your age", it can automatically answer the question "how old are you" because the two questions are considered semantically similar.

However, an AI chatbot must decide what it should automatically answer and what it should not. Consider Kate, a first time visitor to Sarah's website. She asked "where do I sign up". Since Juji (Sarah's chatbot) could not find a closely matched question that it could automatically answer, should Juji consider Kate's question the same question as "where do I log in"? Are these two questions close enough to share the same answer? In many cases like this, an AI chatbot faces making a decision under uncertainty. Instead of simply making a decision on its own, the best AI chatbot often involves users in such a decision-making process.

As shown below, the Juji AI chatbot suggests a list of closely related questions to Kate so that she can help determine which might best match with her intention.

From Kate's point of view, she is given the choices to select a matched question and obtain a potential answer to her question. Through the suggested questions, even if Kate could not find a good match, she still gets an opportunity to learn about the capabilities of the chatbot, which in turn facilitates her future communication with the chatbot.

Assume that Kate thought "Where do I log in?" was close enough to her question. She made her choice and was then led to the login page, where it also has a button for creating a new account. Kate's request was fulfilled.

Design considerations for question recommendation

If you decide to power your chatbot with a question-recommendation engine yourself, here are several things to consider:

How often should a chatbot make question recommendations? What is the threshold or condition to trigger the recommendation?
How to determine a list of "similar" questions to recommend? What other criteria beyond semantic similarity should be considered in ranking questions?
How to present the top-N matched questions? How many should be presented (what is the N)?

The first aspect is to consider how often a chatbot should make a suggestion. Ideally, a chatbot should only make question recommendations when it is necessary. If it recommends too often, a user will feel that a chatbot has little intelligence and the process is too burdensome. On the other hand, if a chatbot makes too few recommendations, it may annoy users with wrong or no answers to their questions.

The second aspect is to consider what to recommend by balancing a set of criteria. An "optimal" list of question suggestions should be closely related to a user question while covering diverse topics to avoid "tunnel vision".

The third aspect is to consider how many questions to suggest. If there are too few, a user may not find a matched question; if there are too many, a user may feel overwhelmed.

All these considerations above are to maximize the probability for a user to find the information s/he seeks while minimizing the user's effort/burden in this process.

As you can see, just like building any recommendation engine, supporting effective question recommendation is non-trivial. The good news is that if you don't wish to do it on your own, you can always utilize what's already built, like the built-in question recommendation capabilities automatically provided by Juji.

Although no chatbots are perfect, there are always ways to improve a chatbot and make it serve their customers better. In the worst case scenario, if users still don't find the suggested questions helpful, what should a chatbot do or what should a chatbot designer do? In my next blog, I will talk about how to help chatbot designers track unanswered user questions in real time so their chatbots can be improved in real time!

In the meantime, I'd like to invite you to sign up free and follow this short tutorial to create your own customer service chatbot that can answer users' free-text questions and gracefully handle the unknown ones by suggesting relevant questions.

Chatbot as a Career Counselor? What do Students Think?

2020-05-28T16:54:43Z

Challenges

Students have difficulty receiving career advice due to the low advisor-to-student ratio. I know that's true for me. Sometimes I would forgot to make an appointment with a counselor, dropped-in, and waited an hour to see one. Other times I avoided the office entirely because I knew appointments were full. How could such issues be solved? Maybe it's possible to create a chatbot that could act as a career counselor.

The implications of a career counseling bot could be significant, it could reduce the impacted schedules of counselors by augmenting their work and answering student inquiries 24/7. However, understanding whether chatbots could act as a career counselor is still under studied.

To gain a better understanding of the possibility, I assisted in designing and conducting a research study where a chatbot made on the Juji platform was used to survey students. The bot asked 350 students in the Lucas College of Business at San Jose State University on their thoughts of how it could serve as a counselor.

Findings

The business students were very creative and gave interesting insights on chatbots as career counselors. Students' ideas and requirements for a career counseling chatbot were summed up into 4 major themes:

Offering information and recommendations
Providing intervention on career development
Augmenting counselors' work
Providing career counseling

First, students would like a career counseling bot that can offer information and recommendations. More specifically, they would like to get answers to questions related to the university, such as financial aid deadlines, school holidays, and career advice. They also have high hopes for the chatbot and would like it to recommend career choices based on their conversation with the chatbot. In other words, such a chatbot should be able to capture students' personality, likes, and dislikes, and match them with a career best-suited for them.

Next, providing intervention on career development was another theme commonly mentioned by students. Interestingly enough, students wanted varying degrees of intervention in their career development. The first intervention method they suggested was reminders for upcoming career events, such as career fairs or career workshops, like resume building workshops or LinkedIn profile workshops. The next intervention method was goal tracking. Students wanted to be able to set personal goals and deadlines for themselves through conversation with the chatbot and receive reminders on their progress towards those goals, such as receiving an internship. The final intervention method was assistance in career development through mock interviews and resume scanning. Notice that the degree of intervention becomes increasingly involved.

Another major theme was augmenting counselors' work. By this, students meant that the chatbot could act as a medium between advisors and students, connecting the two parties together. They suggested that the chatbot should be able to answer questions automatically or redirect users to a human to answer their questions. Another common suggestion was gathering students' concerns before an appointment with a counselor and creating those appointments for them.

The final theme was providing career counseling. In other words, students wanted the chatbot to become the counselor by providing career advice and giving guidance on the necessary courses for graduation.

The four themes derived from the study can be used to guide the implementation of a chatbot as a career counselor for students. For more detailed information on this study, please refer to our full paper published in The Americas Conference on Information Systems 2019 (AMCIS 2019).

A chatbot as a counselor could be helpful for both students and counselors. For students, they could receive advice and/or answers to burning questions, 24x7. For counselors, the repetitive tasks, such as answering frequently asked questions could be done by the chatbot, leaving them more time for more important work, such as giving advice. Overall, I think this use case could be extremely valuable for chatbot designers or developers who focus on education applications.

How to teach your chatbot to answer complex user questions

2020-05-11T01:07:45Z

If you are creating a chatbot to automate customer support and handle customer questions, you could do so easily on Juji by preparing a question-answer (Q&A) list and uploading the list to enable free-text Q&A instantly.

While the majority of customer questions can be handled in such a way, there are cases that a chatbot needs to obtain additional information from users before producing a satisfactory answer. Assume that a user asks "Where is the nearest branch?" and your business has multiple branches. A good AI chatbot should first ask where the user is located and then respond with the location of the nearest branch.

I will now show you how this can be easily done using the Q&A Board in Juji Studio.

In addition to providing a simple answer to a user's free-text question, I can use Juji to add an extended or multi-step answer to a user question. As shown above, just click on the extended reply button in the answer column. It will open up a panel for me to create a new chat flow as the answer.

As shown below, the flow editor works just like the main chat flow editor in Juji Studio, where I can create custom topics or reuse already created topics in the main chat flow.

Using the example I mentioned above, I first create a topic that asks which city a user lives. Then, I customize the chatbot actions to respond with different locations nearest to the users based on their answers. Once I am done with the flow, I go back to the Q&A Board to submit the changes.

It's that simple. Let's see the effects.

As shown in the screenshot above, not only can such an AI chatbot provide the nearest branch's location according to a user's response, but it can also remember and resume the original conversation flow after it finishes the Q&A flow. In fact, Juji chatbots automatically handle such a side Q&A flow anytime during a conversation and gracefully resume to their original task flow (e.g., the workflow of making a reservation).

Such an extended Q&A flow is very powerful and can be used in many applications and domains. Below I share several more use cases.

Healthcare

In patient engagement, as shown below, a patient is requesting to see a doctor. A healthcare chatbot can ask the patient about his/her symptoms before giving a proper response.

Office Administration: Making a Booking

Assume that a chatbot is helping a customer book a table at a restaurant. As shown below, during the booking flow, a customer interrupts the flow with a question about the use of a coupon, as shown below. To respond properly, the chatbot activates a Q&A flow before returning to the table reservation flow.

As you can imagine, such a Q&A flow can be used in many similar chatbot tasks, such as booking a travel reservation, booking an appointment with a car salesperson, booking a house tour with a real-estate agent, or even booking a webinar event.

E-Commerce

Similarly, in e-commerce, when a user asks for a product recommendation, a smart retail chatbot asks the user his/her preferences before making a suggestion.

To help chatbot designers and developers build the best AI chatbots, Juji has made the handling of complex user questions super easy. One just needs to define a multi-step Q&A flow, similar to defining a main chat flow. A Q&A flow also functions like the main chat flow, supporting all the powerful functions, such as detecting user sentiment and personalizing messages based on user characteristics. The only difference is that a multi-step Q&A flow can be activated anytime during a conversation to provide users with just-in-time help.

If you want to see more of such powerful Q&A response in action, check out this YouTube video. The Q&A Board also provides real-time notifications, see how you can continue to evolve your powerful chatbot.

DIY Your First AI Chatbot: If You Can Use PowerPoint, You Can Build an AI Chatbot

2020-04-15T23:18:27Z

Gartner predicts that 80% of business applications will have a chatbot by 2021.

"Building a chatbot seems complicated, I don't really have the time to learn and build one."

"I want a chatbot for my business but I don't have the technical skills to build one."

"I can't code so I probably won't be able to build a chatbot on my own."

Learning something new can be intimidating. If you have some of the thoughts mentioned above, this blog post may completely change your thinking and leave you with a new perspective on chatbot making.

Thanks to great advances in technologies, now you can use a user-friendly, but also super powerful Do-It-Yourself (DIY) chatbot builder to make custom, capable AI chatbots without coding and within minutes.

Let me show you how you can quickly DIY your first AI chatbot on Juji to send custom messages and support free-text question and answering (Q&A).

......

Template Selection

Juji provides a set of chatbot templates. For our purposes, let's choose the blank template that we can easily customize.

......

Customize Chatbot Welcome and Wrap-up Messages

The blank template comes with two conversation topics, one is a "Welcome" topic that greets users at the beginning of a chat, while the "Wrap-up" topic lets users know if the chatbot will hang around or greet a user goodbye. As shown in the screenshots below, you can easily customize both the welcome and wrap-up messages to suit your audience. If you prefer, you can also customize the ending options, such as looping back to the beginning of the chat.

......

Support Free-Text Q&A

One of the key functions of an AI chatbot is its abilities to address user's natural language inquiries during a chat. To do so, you can teach your chatbot with a list of Q&As in a CSV file. First, you can download the CSV file template from the Q&A dashboard as shown below.

As shown below, the CSV template contains four columns:

ID | Question | Answer | Comment

You can give a meaningful ID to index your Q&A pair and also use it to group alternative question or answer expressions. With the same ID, you can have multiple question variations and one answer for all those questions. Every chatbot template automatically comes with a default Q&A list, which you can customize. Pay attention to the answers in the CSV ended with "#", indicating a default system reply to be modified. Here is more information on customizing Q&As.

Once you are done with your Q&A list, save and upload the CSV file on the Q&A dashboard. Now you can preview your chatbot and test out the Q&As you just put in.

Juji uses machine learning to match a user-inquiry with the questions you entered by their semantics. In other words, Juji can recognize what a user is trying to convey. As a result, you will notice the following chatbot behavior. First, even if your questions don't exactly match the ones you entered in the CSV, your chatbot can still answer the questions correctly. Second, if your chatbot is unsure about a match, it will then recommend possible questions, giving users a chance to select the correctly matched question as shown below.

Once you are satisfied with your chatbot, you can then deploy it to meet your audience, either on your Facebook page or your website.

This Youtube video recaps what we covered in the blog post. As you can see, creating an AI chatbot is easy and fast, perhaps easier than making PowerPoint slides.

If you wish to learn more about AI chatbots and how to build one, please meet us on this sub-reddit or follow us on Juji Linkedin page.

Juji Offers Free Chatbot Services to Healthcare Institutions

2020-04-09T20:42:31Z

During this difficult time of COVID-19 pandemic, healthcare organizations and medical institutions are inundated with requests and demands from the general public. As an AI chatbot company based in Silicon Valley, Juji would like to lend them a helping hand. We are currently offering our chatbot services free-of-charge to these institutions.

Juji's platform automatically answers user questions in Facebook Messenger or on a website based on a curated list of questions and answers. For example, a group of volunteers from the New Voices Program of the National Academies of Sciences, Engineering, and Medicine have used Juji to create an AI chatbot, Jennifer, to clarify misinformation and answer public's questions about COVID-19.

We are offering the same free service to all health organizations or medical institutions, so that critical health information can reach more people in this time of need. With extended home stays, online media becomes a significant source of information. We can help set up a free chatbot in Facebook Messenger associated with public Facebook Page or a chatbot on the institution's Website.

Setting up a chatbot using the Juji platform is fast and easy, requires minimal technical expertise from the users. For instance, the National Alliance on Mental Illness - Long Beach Area has added a Juji powered AI chatbot, Julie, on their homepage to help answer questions. The work was completed in a single day under Juji's guidance, and was done without involving other IT personnels.

If you are reading this, please forward this information to any relevant institutions, so that this new and free automated information channel becomes accessible to more people. The adoption of an AI chatbot will help alleviate any manpower shortages while reserving precious human resources for more urgent matters, e.g., coordinating logistics and caring for patients. The time to act is now.

This Is How We Revamped the UI in Less Than a Month

2020-04-08T06:33:36Z

We all know that user interface (UI) development is an iterative process. It is important that we can iterate quickly based on user feedbacks. At Juji, we have been constantly searching for solutions that enable faster iterations for our Juji Studio product. Around the end of last year, we did a major revamp of Juji Studio UI. By all accounts, this change made a huge difference in term of usability of Juji Studio. More importantly, we can now iterate much faster than previously possible. What's more, we did the wholesale changes in less than one month! Here is how we did it.

One of the main innovations that we came up with is a new way to handle the client-server communication for a single page application (SPA). Instead of sending resource oriented API calls, the client sends the changes of its data model back to the server, and server applies the changes to its own copy of the data model for persistence. We developed the necessary diffing library and open-sourced it on github.

For some background, we have long adopted GraphQL as the API language, with which the server side and the Web client speak to each other. GraphQL has been a good choice, as it is fairly flexible and there are a slew of tools to make it convenient to use. As our product grew in features, we started to feel that changes became harder to make, and iterations slowed down to a crawl.

Juji Studio interfaces with many different stateful systems to enable DIY AI chatbot creation. For example, it needs to work with our chatbot DSL generator to generate the code from the UI, communicate with the chatbot runtime system in order to enable preview of the chatbots, talk to deep learning compute services to get recommended topics, handle long running analytical jobs by using a Kafka cluster, and additionally talk to various databases systems to query and transact information. All these backend work need to be surfaced to the UI in various ways to give user a smooth experience. Not surprisingly, our GraphQL schema file alone grew to be over 2K lines of Clojure EDN code. For a language famous for its conciseness, this is a huge amount of code just for describing the API schema.

We started to question: did we really need this many API calls? Was there a way to cut the number down? Stepped back and examined what we were doing, we decided that this was not sustainable. What we were doing is the standard way of building APIs: viewing the client-server communication as retrieval and manipulation of remote resources. Even with GraphQL, we are still thinking in the REST way, or the CRUD way. For any resource, one would need at least four API calls, one to create it, one to query it, one to update it, and one to delete it. However, unlike many CRUD applications that only need a database backend, the GraphQL API of Juji Studio cannot be automatically generated from a database schema because of the diverse systems that it interacts with, so we wrote many repetitive APIs for various different things.

The worst part was not the number of repetitive APIs, but the scatteredness of business logic. To change anything, we needed to update the database schema, update the backend code, update the API schema, update the service side of the API, then update the client side of the API, and finally, we got to write UI code for the change that a user would actually see. This is a very long chain of events, and things could break down anywhere. Particularly, as it was often the case, these different links of the chain were handled by different developers. With the added communication overhead, it was a miracle that we could get anything changed at all!

In retrospect, the standard resource-centric view of software architecture does not really play well with the concept of SPA , where the bulk of the business logic expects to happen in the rich client. In actuality, the server side mostly serves as the persistence layer of the said business logic, and should not be concerned with its details. For our use case of DIY AI chatbot creation, most of the complexity are concentrated in the user interfaces. Forcing a client-server functionality split only serves to increase incidental complexity.

With this clarity, it is now natural to arrive at what we are doing right now:

The Web client works on a data model in the form of an EDN document, and does whatever manipulations of the document necessary to serve the user.
Periodically, the client sends to the server a diff between its current state of the document with its previous state.
The server knows nothing about what the client did, it just blindly apply the diff to its own copy of the document, persist the new document, and sends the document SHA back to the client.
The client compares the incoming SHA with its own to know that its state has been successfully persisted; or it will send the whole document back to overwrite the server document.

With this scheme, to change the functionality of the UI, all that is necessary is for the stake-holders to have a discussion with the frontend developers and the UX designer. Backend developer involvement is not necessary. Faster UI iteration is easily achieved. As to the APIs? Many of the original API calls are now removed, replaced by the three generic fetch-doc, send-diff, and overwrite-doc calls, which the frontend developers do not touch once they are written.

Because the frontend developers had complete control of the data model and did not need to be concerned with the server side, they could quickly revamp the entire UI of Juji Studio. It took less than a month to go from the design conception to the end of implementation. And the resulting product is nothing short of stunning.

Obviously, I have glossed over many implementation details of this data diff driven architecture. For example, issues about data validation, dealing with errors, dealing with multiple browser tabs and multiple browsers, and so on, are too numerous for the space of an article. The gist of this post, however, is to introduce this idea of data diff oriented software architecture, and to outline its benefits of increasing modularity and decoupling, and thereby decreasing the friction for change and the time to market.

Three Big Ideas Behind the Surprising Power and Ease of Use of Juji Platform

2020-04-07T01:10:00Z

When people first use Juji, they are often amazed by how easy it is to create an intelligent chatbot with the platform. This reaction of pleasant surprise is particularly pronounced for people in the know, i.e. technical people who have actually done relevant work before. I am talking about the CTOs, the NLP researchers, and the employees of big technology firms.

A fun anecdote: our first Facebook ads a couple weeks ago was rejected due to "Unacceptable Business Practices", even though the ads was just a screencast of a user using Juji to create a chatbot. Obviously they reverted the decision after I complained, but it did indicate that what Juji offers is often considered too good to be true.

I give full credit to our wonderful engineers and designers, whose ingenuity and heroic effort makes Juji possible, and to my partner Michelle, whose relentless advocacy for users drove us here. On the other hand, there might also be a few things that I did not screw up completely as the architect. In retrospect, here are three high level ideas that might have helped, or in marketing speak, "three big ideas that make Juji stand out".

To give more context, my point of comparison is regard to the chatbot platforms that truly have elements of artificial intelligence, not the numerous button bot platforms that offers no more than a graphic user interface in a message box. In this regard, I am mainly comparing Juji with chatbot offerings from a few major technology companies: Amazon Lex, Facebook Messenger, Google DialogFlow, IBM Waston, Microsoft Bot Framework. If you insist on knowing my opinions on these platforms, my ranking of them in term of technical capability is the following: IBM > Microsoft > Google = Amazon > Facebook. Obviously, Juji is at a mile above them all ;-).

1. Agency

Call me crazy, but I actually want to build artificial general intelligence (AGI), someday. Of course, I am not crazy enough to bet a company on AGI. However, I do believe that AGI could be reached incrementally, by building smaller and practical intelligent things one at a time. Just like how humans build anything impressive, the practices will always be ahead of the big and beautiful theories. Chatbots surly sound like one of those smaller and practical goal posts that might lead to something big in the future. All these is to say, I want my chatbot platform to have the potentials to grow into something more ambitious, perhaps an operating system for intelligence?

The first technical capability necessary for intelligence is what I call agency. Animals have agency because they act on their own, whereas objects such as stones, cups, or keyboards do not. Objects do not have agency because they only react, but do not act on their own. Obviously, there is no possibility of intelligence if there is no agency. In term of computer software, for a platform to support building software components that have agency, it must allow the components to have their own execution loops.

There is an important distinction between bots that just respond to user input or external events and those who run their own execution loops. The former share a system event loop and is reactive, and the later let each agent to have its own execution loop, hence can potentially be proactive, i.e. acting on their own or have their own mind, so to speak. It is the later that should be regarded as real agent.

Among the major chatbot offerings above, only IBM Waston bots seem to have agency. The rest of the offerings are either deployed as Web hooks (DialogFlow, Lex), or run as callback handlers (Bot Framework). These are all reactive bots that cannot act on their own. In Waston bot, the dialog is defined as trees of nodes, where each node is a production rule (i.e. If-Then). The developer does not control the execution of the dialog, but the agent itself runs an execution loop to go through these trees and act accordingly. In principle, this enables the agent to be proactive. I do not know if IBM Waston actually does this proactive firing of rules in practice, but this is precisely how Juji bot works.

When an end user comes to one of the Juji chatbot deployment on either Facebook or the Web, a new Juji bot is instantiated on the spot for this encounter. Each Juji bot runs two execution loops simultaneously, one reactive and another proactive. The end result is that the bot may speak any time on its own, not just react to user input. It is also easy to keep an Juji bot instance running indefinitely, act on its own to proactively send information out to a user, based on a schedule, or some environmental contexts that the bot creator has determined. Essentially, each end user gets her own unique Juji bot, who overtime would potentially develop a unique relationship with. This is a far cry from the universal bots in your living room now that are not personalized and are only reactive.

2. Topic Abstraction

The second technical aspect I care about is the abstraction of conversation embedded in the system. One factor I look at is the concept of natural language understanding (NLU) used. The majority of NLU systems are modeled on a concept of intent, referring to what user wants to do. This reflects a fundamental bias of these systems that were originated from academic research, where the narrow goal that the academics have set themselves up is to help users to accomplish certain tasks, hence it is central to understand the intent of user utterances.

I regard the reliance on intent as a severe limitation, because human-bot conversation may not be about user’s intent at all. For example, what about the bot’s intent? Considering only user’s intent limits the application of bots to some boring application domains such as customer support, question answering, internet of things, or e-commerce, where user ask questions that bots try to answer or speak their wishes that bots try to fulfill.

For more interesting applications such as marketing research, job interviews, gaming characters, customer on-boarding, educational companion, mental health assistant, and so on, the chatbots need to have their own agenda, which the often used intent concept simply does not cover. Among the major systems, the only exception is Microsoft BotKit, where intent is not explicitly hard coded in the system. On this front, Juji goes a step further, Juji bot can have complex and explicit agenda that go beyond either user or bot’s simple intent. These agenda are also present in Juji's question answering system, so a user question may lead to a completely different agenda, hence a completely different conversation.

Another factor in the conversation abstraction is the unit of dialog considered in the system. Most systems treat a turn as the basic unit of conversation. This is too granular, because the developers of the bot then have to think of all the possible user utterances at each turn and respond to them accordingly. This is a task not very suitable for a developer to do and the system should give them as much help as possible. BotKit is again the only solution that introduces a higher level concept. BotKit has a concept of a thread, which handle a sequence of turns. However, this is not good enough, because a thread can only be executed, and the only thing one can do with them is to jump among threads.

Juji’s abstraction is called a topic, which may have zero, one or multiple turns in them. Most importantly, topic is the first class citizen in Juji platform, where one can create a topic on the fly, pass arguments to it, pass a topic around, look up a topic, and do all kinds of things with them. This flexibility enables Juji to supply a large library of reusable mini-conversations (represented as topics of course) that users can simply compose into a full bot. Developers are largely alleviated from the burden of trying to anticipate user’s next input, because Juji has many reusable topics that handle all kinds of user digressions and dis-behaviors that a bot developer is not well equipped to anticipate.

The topic abstraction also enables easy representation of complex conversational logics and contexts as plain data structures. This data centric view leads to pervasive code generation throughout our system. Some data in, other data out. Everything flows as data and can be generated on the fly. Such maximum flexibility makes it easy to create an easy to use chatbot creation user interface, without forcing users to learn strange NLP jargons, such as intents, entities, slots, and so on.

3. Symbolics as the bones and ML as the flesh

With the popularity of deep learning (DL) and machine learning (ML) technology, it is not surprising that most of the systems listed above have natural language processing (NLP) capabilities based on them. These capabilities are the must haves in an AI chatbot platform if the tasks require understanding free-text utterances. Although most button bot platforms on the market today do not offer any NLP capabilities, I expect some of them will integrate them eventually. However, in my opinion, pursuing competitive advantage in raw NLP model performance alone is rather a futile exercise, because these technology are rapidly commoditized and the differences among vendors are minimum.

The real competitive advantage is the ease and speed with which a new NLP model can be deployed in production. Here DialogFlow and Lex seems to be very capable, as they are essentially plumbing mechanism of data flows, so new NLP models should be easy to be plugged in. The NLP integration story of Waston and Facebook (wit.ai) is not as clear because NLP capability seems to be part of the system, which is actually a weakness, because one wants to iterate on these often and fast.

Juji takes a unique and practical hybrid approach to integrate DL and ML based NLP with the so called traditional AI approach. As described in my 2018 talk, our slogan for the integration is "Symbolics as the bones, and Machine Learning as the flesh". Consequently, Juji has a complete story on NLP integration. One can either use Juji’s built-in NLP models, run one’s own code in Juji’s sandbox, or call out to third party code easily. The components are loosely coupled, yet are all within the comfort and convenience of a single system.

In summary, all three big ideas fit together nicely to create the unique Juji chatbot creation experience that is both easy and powerful. For example, we can easily integrate any end-to-end DL based conversational techniques to support lively chitchats. At the same time, because our customers have placed a high emphasis on controllability of the bot utterances, our system can easily control the flow of the conversation and leads it naturally towards customer's business goals. We are able to accommodate these seemly conflicting requirements, because we have set out to design a system that produces true agents, has the right level of abstraction and is practical rather than ideological about how to achieve AI.

How to build a Smart Chatbot in a Few Minutes to Answer Free-text Questions

2020-02-20T04:44:00Z

People always have questions. You are running a business, you find yourself keep answering customer questions. You are a community manager, you find yourself keep answering community member questions. You are on Facebook Messenger, you find yourself keep answering your friends’ or colleagues’ questions. Sometimes you just want someone or something to answer these questions for you!

Chatbots come to the rescue. Here I want to show you how to build a smart chatbot in minutes to answer free-text questions. Your chatbot can answer questions specified in diverse forms. It can suggest answers even if it is unsure what a user is asking. Moreover, it automatically tracks unanswered user questions, so you can supply answers to these new questions and improve your chatbot quickly.

Let’s use the Juji platform (https://juji.io) for this task. Once you log in, click on “+ AI Helper” to create a chatbot. For this exercise, let’s use a blank template as shown below.

Then, click on “Customize Chatbot” and then go to the Q&A BOARD directly.

First, click on the “Download” button to download the CSV template. All you need to do is to fill in your Questions and Answers as following:

The ID column is useful for you to group the alternative expressions of a question so they can share the same answer(s), although the chatbot can also automatically detect many matched question expressions. A comment is optional. You can learn more about the format here.

Once you are done, save it into a csv file (simply choose to “Save As…” a .csv in Excel) and then upload it.

Now the moment of truth! You can always PREVIEW your chatbot before publishing it. Try asking questions in different ways to test the power of AI. If you wish to include more questions, edit the CSV file and upload it again.

Once you are ready to publish, click on DEPLOY to publish your chatbot to a Facebook Messenger or on a website. More details here.

After your chatbot is published, it’s always good to go back to your Q&A BOARD and check if there are any questions your chatbot could not answer or was not sure about the answer.

As shown above, there are two unanswered questions. One has a chatbot-suggested answer highlighted in red. One requires an answer since a good match could not be found. I can enter or edit the answers and then submit them. Once the answers are submitted, the chatbot learns instantly and now can answer these questions.

The Juji platform is a one-stop-shop for building smart chatbots. It allows you to easily create a chatbot to answer questions anytime during a chat and still remember where the conversation is! (check out this YouTube video and this live gallery of smart Juji chatbots).

Remember to check your Q&A BOARD regularly and update your chatbot with new answers. This way, your chatbot keeps learning and becomes smarter and smarter.

Have fun building chatbots and may the chatbot force be with you!

Chatbot Tsunami: The Good & Bad

2020-01-28T21:58:43Z

The widespread use of cell phones and social media has made text-based communication, also known as texting, a mainstream communication method. For example, a Gallup polls hows that texting is the dominant communication method for Americans under 50. Popular text messaging platforms, such as Facebook Messenger and WeChat, all boasted over 1 billion active users monthly.

However, a person can hardly text with multiple parties at the same time, let alone do texting 24/7. That’s where chatbots come to the rescue. A chatbot engages a person in one-on-one text messaging to automate various tasks. For example, Sephora uses a Facebook Messenger chatbot to help users book an in-store makeover, while Kohl’s uses a chatbot to help customers locate a store or find deals.

As a marketing expert pointed out, a chatbot is meant to scale out brand-customer communications and assist customers better. First, a chatbot will never feel tired and can text with hundreds of thousands of users simultaneously to provide them with immediate help. Second, a chatbot will always remain calm and polite to help users even in difficult, emotional situations. Third, a chatbot does not have a personal agenda or ulterior motives and could offer users with objective and fair services.

On the other hand, studies show that most chatbots are just bots with little chatting capabilities. For example, one study shows that 70% of Facebook Messenger chatbots were unable to answer simple questions, and another study shows that it is still quite challenging for chatbots to handle a simple task, like scheduling a meeting. Additionally, these limitations cause “chatbot spam” — using chatbots to simply broadcast one-way messages, similar to email spam, but in the form of text messaging.

Although it may take 10 to 500 years to achieve the ultimate human-machine symbiosis envisioned by Licklider, a practical and urgent question seems:

What can we do today to take full advantage of the good characteristics of chatbots and avoid their bad ones?

My next post, AI Chatbots: Their Good Uses, will attempt to answer the question.

2026-02-16T05:43:31Z

google-site-verification: google87579b2b505115de.html

Juji Open Source Software

2022-05-13T05:24:53Z

Juji Open Source Software

Datalevin: a fast Datalog database and search engine.
Editscript: a library to diff and patch Clojure/Clojurescript data structures.
symspell-clj: a library to spell-check text.
Gibberish-detector: a library to detect gibberish user input.
Juji cli-client: a node.js command line client that demonstrates the use of Juji API.
Juji Speak: a GUI client that uses Juji API along with voice API (STT and TTS) to power a Juji cognitive AI assistant with voice.

Improve Online Learning Success with Cognitive AI

2020-10-28T05:24:53Z

Click on me to chat with me

Improve Online Learning Success with Cognitive AI

2020-10-28T05:24:53Z

Click on me to chat with me

Challenges and Opportunities in Online Learning:
Four Key Findings from University Leaders

Our interviews with 42 university leaders revealed the insights into managing online programs. Read the abstract below or chat with our AI to access the report.

image by Christin Hume

Abstract

Our report outlines four key findings from interviews with 42 leaders of 33 online academic programs across 28 U.S. universities nationwide during the summer of 2021. The findings uncover the insights of building and managing online or e-learning programs at higher education institutions, including main challenges and expectations.

These findings inform education technology providers as well as university decision makers on developing and selecting suitable technologies especially AI-powered solutions that can improve student e-learning experience and success at scale.

Challenges

participants voiced top-3 challenges in managing online programs: keeping students engaged (61%), recruiting prospective students (45%), and developing quality online courses (35%).
The top priorities for the participants to address are: student engagement (39%), student recruitment (23%), and course development (22%)
The current solutions to the top priorities are mainly human-driven and difficult to scale, while technology-driven solutions are less than satisfactory.

Opportunities

University decision makers expect technology guidance and easy-to-adopt technologies that can better support online education success.
Two AI solutions could help address universities’ top priorities: (a) AI assistants (chatbots) to personalize student engagement at scale, and (b) content-based AI technologies to personalize online teaching at scale.
Three technology must-haves, quality of services, integrability, and usability, can be used to guide the development, evaluation, and adoption of technologies for online programs.

Improve Employee Engagement with Cognitive AI

2020-10-28T05:24:53Z

image by Tim Mossholder

Opportunities

University decision makers expect technology guidance and easy-to-adopt technologies that can better support online education success.
Two AI solutions could help address universities’ top priorities: (a) AI assistants (chatbots) to personalize student engagement at scale, and (b) content-based AI technologies to personalize online teaching at scale.
Three technology must-haves, quality of services, integrability, and usability, can be used to guide the development, evaluation, and adoption of technologies for online programs.

Click on me to chat with me

Why Juji is the best AI chatbot platform

2020-03-31T05:24:53Z

The Easiest Way to AI Chatbots

Out-of-the-box AI for instant gratification

AI CHATBOTS FOR

Business Owners

automate customer interaction with a human touch

Augment Manpower. Scale Business. Grow Profit.

better customer experience

24x7 Availability. Personalized Care. Brand Bonding.

USE CASES

JUJI TOOLS

JUJI ADVANTAGES

Juji offers the world's best AI chatbots. They enable you to automate customer interactions and achieve maximum ROI.

QUICK TEST

AI CHATBOTS FOR

Freelancers | Agencies

JUJI ADVANTAGES

Juji offers a one-stop-shop to build, deploy, and manage AI chatbots. Easier than making non-AI chatbots.

QUICK TEST

power chatbots with AI to up-sell chatbot services

Superior chatbots for handling complex tasks. Anytime Q&A. Personable.

maximize profit margin

Serve more powerful chatbots, faster. Improve the bottom line.

USE CASES

JUJI TOOLS

AI CHATBOTS FOR

Marketing Professionals

run conversational marketing without bothering developers

Full control with no coding. Faster results. Accomplish more.

more time on high-value tasks

Automate repetitive tasks with chatbots. More time on fun, creative tasks.

USE CASES

JUJI TOOLS

JUJI ADVANTAGES

The Juji platform enables marketing team to DIY powerful AI chatbots with no coding and no dependencies on developers.

QUICK TEST

AI CHATBOTS FOR

Researchers

JUJI ADVANTAGES

Juji's DIY platform enables you to easily build, deploy, and monitor chatbot behavior and study their impact.

QUICK TEST

chatbot is the new social media

67% customers used chatbots to interact with brands. 80% businesses will use chatbots by 2021.

easiest way to conduct chatbot research

Design, build, and deploy. Experiment with different chatbot behavior. Study their impact.

USE CASES

JUJI TOOLS

The best AI chatbots for information elicitation

2020-03-31T05:24:53Z

JUJI SOLUTION

Cognitive AI for Information Elicitation

Juji powers cognitive AI assistants in the form of chatbots to engage with each respondent in a two-way conversational survey or interview.

Elicit higher-quality responses and make happier respondents.

LEARN MORE

SOLUTION OVERVIEW

WHY JUJI'S BETTER

HOW IT WORKS

CASE STUDY

GET STARTED FREE

Cognitive AI Survey Assistant

AI assistants with human soft skills serve as empathetic and responsible interviewers

Many organizations, such as educational institutes, healthcare organizations, and government agencies, conduct surveys or interviews to gather information from target audiences. One of the main challenges is how to overcome survey fatigue and gather quality responses.

Juji's cognitive AI assistants can augment researchers, scientists, and business professionals alike to engage with respondents in a two-way conversation at scale, delivering engaging survey experience and eliciting high-quality responses. Follow the design best practice to maximize conversational survey outcomes.

AI Assistant for
Conversational Surveys

Engage with respondents in two-way conversations, asking them questions and eliciting their responses.

Main Responsibilities

ask respondents open-ended and choice questions to elicit both qualitative and quantitative input.
answer respondent questions during a survey to make clarifications, provide guidance, or prevent faking.
read between the lines to infer audience psychographics and better understand the "why" behind the collected responses.
automatically summarize free-text responses to open-ended questions to generate consumable insights

CASE STUDY

Best Practice

The best survey chatbot should engage survey respondents in a true two-way conversation to deliver an engaging experience and elicit quality responses. With Juji Studio, you can easily design a survey chatbot to achieve all the following and conduct effective surveys:

ask a mixed set of multiple choice and open-ended questions to elicit both quantitative and qualitative insights
anticipate and address potential user questions, including clarification questions and user excuses.
use Juji pre-built AI to make a conversational survey engaging, such as paraphrasing a question and personalize a request.

CREATE MY OWN

Why Juji is Better

ACHIEVING THE BEST ROI

Online surveys are important for conducting user or market research. However, survey fatigue is now a major challenge faced by anyone who wishes to collect data due to extensive use of online surveys.

Juji's conversational survey solution humanizes surveys with AI to combat survey fatigue and help you achieve the best survey ROI from two aspects: better survey results and faster to insights. No technical skills required.

Better Survey Results

Compared to traditional online survey tools or other survey chatbots, Juji chatbots can better engage with respondents through a natural conversation and gather higher quality responses.

In particular, Juji chatbots can achieve:

78% better data quality

than traditional online surveys by asking open-ended questions and their follow-ups to elicit rich participant input.

2X completion rate

of traditional online surveys by auto-handling user digressions (e.g., gibberish input and excuses) to ensure survey completion.

Faster to Insights

Juji's pre-built AI and no-code chatbot builder enables you to build and launch high-performance survey chatbots and obtain insights, fast.

100% no coding

similar to using SurveyMonkey or Qualtrics, but with full AI powers to obtain higher quality insights

instant insights

obtaining real-time response (including free text responses) and audience analytics

NO IT RESOURCES REQUIRED.

How It Works

THREE-STEP SET UP
DASHBOARD FOR FAST INSIGHTS

Three Steps to Set up

Juji Studio enables you to build and launch an AI chatbot for conducting conversational surveys in three (3) quick steps with no-coding.

LEARN MORE

STEP 1
Enter Survey Questions

Like creating a survey on SurveyMonkey, create a chatbot by entering multiple choice and open-ended questions. Customize each question as needed, e.g., indicating required questions or user response length.

STEP 2
Test Conversations

Preview your chatbot often for fast testing and improvement. Follow these chatbot design tips to ensure the quality of your chatbot, e.g., adding answers to anticipated user clarification questions.

STEP 3
Launch Chatbot

Once you're ready, you can launch your survey chatbot. You can launch it on your website, Facebook Messenger, or your mobile app. You can also share the chatbot URL via email or other channels.

GET STARTED NOW

Real-time Chatbot Analytics Dashboard for Instant Insights

Juji Studio provides you with a real-time dashboard to view and interact with survey results and analytics as surveys are being conducted.

You can also download the results in a portable format (CSV) for further analysis or actions.

Choice-Question Response Analysis

For fast comprehension of results, Juji automatically computes the statistics of survey responses collected from choice questions and visualizes the stats.

Free-Text Response Summarization

Juji automatically summarizes the free-text responses to open-ended questions and displays the extracted main themes (topics) and representative statements.

LEARN MORE

Case Study

Collaborating with Interpret, a market research firm specializing in media, entertainment, and technology, Juji helped build a chatbot to interview about 300 gamers and elicit their opinions about game trailers. For comparison purpose, a typical Qualtrics survey was also conducted and the results from both surveys were compared.

Juji

born: July, 2018

Main Skills

I was made to ask open-ended questions and survey gamers about their opinions on two newly released trailers. I was also able to handle various digressions during a survey, such as gibberish responses and user excuses, to ensure survey completion and the quality of responses.

Main Achievements

2X completion rate

of a comparable Qualtrix survey

78% better data

than the Qualtrix survey

READ THE FULL STUDY

Juji Technical Talks

2020-03-31T05:24:53Z

Juji Technical Talks

Huahai Yang. (2023) Generative AI: Past, Present, and Future – A Practitioner's Perspective. Invited Talk, Institute of Psychology, Chinese Academy of Sciences. [slides]
Michelle X. Zhou and Wenxi Chen (2021) Democratizing AI: Creating Cognitive AI Assistants with No Coding. ACM Tech Talks on AI & Machine Learning. April, 2021. Recorded talk on Youtube.
Michelle X. Zhou. (2021) Cognitive AI Assistants: A New Generation of AI Companions. Invited Keynote, The Workshop on Lifelong Learning and Personalization in Long-Term Human-Robot Interaction (LEAP-HRI) at ACM Human-Robot Interaction (HRI) 2021. March, 2021.
Michelle X. Zhou. (2020) Conversational AI Agents That Can Truly Understand and Help Users. ACM RecSys 2020. Keynote
Wenxi Chen. (2020) A Chatbot Platform that Everyone Can Use. Nvidia GPU Technology Conference GTC 2020. [video] [slides]
Huahai Yang. (2020) Datalevin: a Simple Datalog Database for Building Stateful Applications. London Clojurians Meetup. [slides]
Huahai Yang. (2020) Data Diffing Based Software Architecture Patterns. Clojure/north 2020. [slides] [Q&A Video]
Michelle X. Zhou. (2020) You Really Get Me: Conversational AI That Can Truly Understand and Help Users AAAI'2020 WICRS 2020.
Michelle X. Zhou. (2020) Democratizing AI: The Easiest Way to Scale Business with AI Helpers. ACM SF Meetup 2020.
Michelle X. Zhou. (2019) Getting virtually personal: making responsible and empathetic "Her" for everyone. Youtube Video. ACM IUI' 2019 Keynote.
Michelle X. Zhou. (2019) Getting Virtually Personal: Power Conversational AI to Fulfill Tasks and Personalize Chitchat for Real-World Applications.. ACM ICMI' 2019 Tutorial.
Huahai Yang. (2018) A Clojure Fusion of Symbolic and Data Driven AI. Clojure Conj 2018. [slides]

Juji Patents and Scientific Publications

2020-03-31T05:24:53Z

Juji Patents

Michelle X. Zhou and Huahai Yang. Method and System for Creating Interactive Inquiry and Assessment Bots. Patent Number: US 10,635,752
Michelle X. Zhou and Huahai Yang. Method and System for Match People with Choices. Patent Number: US 10,720,081

Juji Scientific Publications

Ziang Xiao, Q. Vera Liao, Michelle Zhou, Tyrone Grandison, and Yunyao Li. 2023. Powering an AI Chatbot with Expert Sourcing to Support Credible Health Information Access. In 28th International Conference on Intelligent User Interfaces (IUI '23), March 27–31, 2023, Sydney, NSW, Australia. ACM, New York, NY, USA 17 Pages.
Michelle X. Zhou, Huahai Yang, Gloria Mark, Mengdie Hu, Jalal Mahumd, and Aditya Pal. 2021. Building Personalized Trust: Discovering What Makes One Trust and Act on Facebook Posts. In ACM Transactions on Social Computing 4, 3, Article 9 (September 2021), 28 pages.
Lina Zhou, Souren Paul, Haluk Demirkan, Lingyao Yuan, Jim Spohrer, Michelle Zhou, and Julie Basu. Intelligence Augmentation: Towards Building Human-Machine Symbiotic Relationship. In AIS Transactions on Human-Computer Interaction, volume 13, issue 2, 2021.
Xu Han, Michelle X. Zhou, Matthew Turner, and Tom Yeh. (2021) Designing Effective Interview Chatbots: Automatic Chatbot Profiling and Design Suggestion Generation for Chatbot Debugging. ACM CHI'2021. Watch a video intro and a recorded presentation.
Ziang Xiao, Michelle X. Zhou, Wenxi Chen, Huahai Yang, and Changyan Chi. (2020) If I Hear You Correctly: Building and Evaluating Interview Chatbots with Active Listening Skills. ACM CHI'2020. https://doi.org/10.1145/3313831.3376131
John E. Wenskovitch, Michelle X. Zhou, Christopher Collins, Remco Chang, Michelle Dowling, Alex Endert, Kai Xu, and Theresa-Marie Rhyne. Putting the "I" in Interaction: Interactive Interfaces Personalized to Individuals. IEEE Computer Graphics and Applications 40(3): 73-82 (2020)
Q. Vera Liao, Yi-Chia Wang, Timothy Bickmore, Pascale Fung, Jonathan Grudin, Zhou Yu, and Michelle X. Zhou. (2019) Human-Agent Communication: Connecting Research and Development in HCI and AI. ACM CSCW'2019 Companion: 122-126.
Michelle X. Zhou. (2019) Getting virtually personal: making responsible and empathetic "Her" for everyone. ACM IUI'2019 Keynote.
Michelle X. Zhou, Wenxi Chen, Ziang Xiao, Huahai Yang, Tracy Chi, and Ransom Williams. (2019) Getting virtually personal: chatbots who actively listen to you and infer your personality. ACM IUI'2019 Companion: 123-124.
Michelle X. Zhou, Gloria Mark, Jingyi Li, and Huahai Yang. (2019) Trusting Virtual Agents: The Effect of Personality. ACM TiiS 9(2-3): 10:1-10:36.
Ziang Xiao, Michelle X. Zhou, and Wat-Tat Fu. (2019) Who should be my teammates: using a conversational agent to understand individuals and help teaming. ACM IUI'2019: 437-447.
Michelle X. Zhou, Carolyn Wang, Gloria Mark, Huahai Yang, and Kevin Xu. (2019) Building Real-World Chatbot Interviewers: Lessons from a Wizard-of-Oz Field Study. ACM IUI'2019 International Workshop on Conversational Agents.
Ziang Xiao, Michelle X. Zhou, Q. Vera Liao, Gloria Mark, Chang Yan Chi, Wenxi Chen, and Huahai Yang. (2019) Tell Me About Yourself: Using an AI-Powered Chatbot to Conduct Conversational Surveys. CoRR abs/1905.10700
Fatema Akbar, Ted Grover, Gloria Mark, and Michelle X. Zhou. (2018) The Effects of Virtual Agents' Characteristics on User Impressions and Language Use. ACM IUI'2018 Companion: 56:1-56:2.
Jingyi Li, Michelle X. Zhou, Huahai Yang, and Gloria Mark. (2017) Confiding in and Listening to Virtual Agents: The Effect of Personality. ACM IUI'2017: 275-286.

Third-Party Publications Using Juji

Fan, J., Sun, T., Liu, J., Zhao, T., Zhang, B., Chen, Z., Glorioso, M., & Hack, E. (2023). How Well Can an AI Chatbot Infer Personality? Examining Psychometric Properties of Machine-inferred Personality Scores. Journal of Applied Psychology, 108(8), 1277–1299.
Hwang, A. and Won, A. IdeaBot: Investigating Social Facilitation in Human-Machine Team Creativity. Proc. of ACM CHI 2021. https://doi.org/10.1145/3411764.3445270.
Yunyao Li, Tyrone Grandison, Patricia Silveyra, Ali Douraghy, Xinyu Guan, Thomas Kieselbach, Chengkai Li and Haiqi Zhang. Jennifer for COVID-19: An NLP-Powered Chatbot Built for the People and by the People to Combat Misinformation. ACL 2020 Workshop NLP-COVID
Sarah Theres Völkel, Renate Haeuslschmid, Anna Werner, Heinrich Hussmann, Andreas Butz. (2020) How to Trick AI: Users' Strategies for Protecting Themselves from Automatic Personality Assessment. ACM CHI' 2020. [video]
Xu Han Am I Asking It Properly?: Designing and Evaluating Interview Chatbots to Improve Elicitation in an Ethical Way. ACM IUI' 2020.
Yu Chen and Timothy Hill. (2020) Teaching Business Students Chatbots: First Forays and Lessons Learned Americas Conference on Information Systems (AMCIS 2020).
Sambhav Gupta, Krithika Jagannath, Nitin Aggarwal, Ramamurti Sridar, Shawn Wilde, and Yu Chen (corresponding author). (2019) Artificially Intelligent (AI) Tutors in the Classroom: A Need Assessment Study of Designing Chatbots to Support Student Learning. Pacific Asia Conference on Information Systems (PACIS 2019).
Terri Lee, Krithika Jagannath, Nitin Aggarwal, Ramamurti Sridar, Shawn Wilde, Timothy Hill, Yu Chen (corresponding author). (2019) Intelligent Career Advisers in Your Pocket? A Need Assessment Study of Chatbots for Student Career Advising. Americas Conference on Information Systems (AMCIS 2019).

The best no code AI chatbot builder with cognitive intelligence

2020-03-31T05:24:53Z

JUJI STUDIO

No-Code Building
Cognitive AI Chatbot

Juji Studio combines cognitive AI with intuitive graphical user interface (GUI), enabling you to build the best AI chatbots without code.

LEARN MORE

WHY JUJI'S BETTER

HOW IT WORKS

GET STARTED FREE

Why Juji is Better

20-50X FASTER, 100% NO CODING

First, Juji enables you to build turnkey, cognitive AI assistants in the form of chatbots faster with out-of-the-box AI. Second, Juji enables you to customize any Juji AI chatbots faster using no-code AI tools. Third, Juji enables you to improve Juji AI faster by supporting real-time AI learning and live chatbot upgrade.

Build Chatbot Faster:
Out-of-the-box AI

Unlike other chatbot builders, with which you have to teach chatbots from scratch, Juji Studio powers chatbots with rich, pre-built cognitive AI powers. So you can focus on designing chat content instead of teaching AI to your chatbot. The ideal conversational AI design setting for beginners.

Here are some of the pre-built cognitive AI powers:

Customize Chatbot Faster:
One-stop-shop, No-code AI

Unlike other chatbot builders, which provide limited or siloed AI tools for you to customize a chatbot, Juji Studio provides integrated, no-code AI tools.

You can pick the tools you need without switching platforms or stitching together multiple AI tools.

natural language processing of user input
message personalization by user persona
support of context-sensitive data queries

Improve Chatbot Faster:
Real-time AI Learning

Few chatbot builders support rapid AI learning, let alone uninterruptive, live AI chatbot upgrade.

Juji Studio enables you to teach a chatbot AI skills (e.g., new chatbot Q&A) and improve a live chatbot instantly without interrupting ongoing conversations.

How It Works

THREE KEY STEPS

First, build an AI chatbot with out-of-the-box AI and no-code AI tools. Second, launch your chatbot to one or more channels. Third, use real-time data insights extracted from the chats to grow your audience and business.

STEP 1

Build

Juji will automatically manage how a dialog flows, handling arbitrary interruptions and maintaining the conversation context.

WATCH AI CHATBOT DEMO VIDEO

Design Main Chat Flow

Define a main flow to describe all chatbot-initiated actions, known as chatbot agenda topics. Like making PPT slides: one slide per topic—conversation topic.

Juji auto-generates a mini-dialog around each topic and links all topics together to form a coherent conversation.

Customize Chatbot Persona

Customize your chatbot persona to match with your brand or tasks. You can select a stock persona with a particular personality or define a custom one. You need a profile photo and a name.

Define Chatbot Fallbacks (Q&A)

User questions or requests can occur anytime during a chat. To respond to such arbitrary user questions and requests, you can set up secondary chatbot actions, known as chatbot fallbacks.

Simply submit a Q&A list in a CSV file or a GUI table to enable chatbot fallbacks.

LEARN MORE

STEP 2

Launch

point-and-click deployment of Facebook Messenger chatbot, website chatbot, or mobile app chatbot, instantly.

worry-free hosting that auto-scales with your business.

Facebook Messenger Chatbots

Deploy your chatbot in Facebook Messenger to automate customer interactions on Facebook. Boost customer engagement, gain deep customer insights, and grow social media ROI.

Website Chatbots

Deploy your chatbot on your website as your "eyes" and "ears" to interact with your site visitors. They can introduce products, answer visitor questions, and gather sales leads.

LEARN MORE

STEP 3

Grow

use chatbot analytics dashboard to view real-time chat stats and audience analytics for the next best action.

integrate with additional tools, e.g., marketing and CRM, to transform data insights into growth and profits.

User Question Analytics

Juji automatically analyzes questions asked by users during a conversation. This informs you what your users care about and what challenges they are facing so you can best help them.

User Response Analytics

Juji automatically summarizes user responses gathered during a conversation. This informs you what your users are like, what motivates them, and what they need so you can best serve them.

LEARN MORE

Chatbot Use Cases: Personalize Human Interactions at Scale

2020-03-31T05:24:53Z

Automate Customer Support

Customer services personalized at scale, 24x7.
Set up in minutes with no coding.

USE CASES

WATCH TUTORIAL VIDEO

Use Case: E-Commerce

My Role

I live in a Facebook Messenger to greet online shoppers, introduce products, and answer questions to convert traffic to sales.

My Superpowers

I work 24x7 and can concurrently help hundreds of thousands website customers—always with a great attitude

Personalize product introductions
Answer product questions anytime
Infer visitor profiles to personalize shopping experience

"This bot was sooo nice. It even complimented almost every time and it was not annoying at all to answer the questions. Very good AI."

—a game shopper

CREATE MY OWN

If you wish to outsource it to an agency, email [email protected]

Use Case: Tech Support

My Role

I live inside a web or mobile application to help visitors navigate information, automate customer support, and infer their needs and wants to optimize customer conversion.

My Superpowers

I work 24x7 and can concurrently guide hundreds of thousands users to the right place at the right time

Lead "lost" users to the right page
Answer user questions
Infer user profiles to personalize engagements

"I just tried Juji. It worked pretty well and I had a nice conversation with it. I was most impressed when it told me at the end I am a very curious person"

—a website visitor

CREATE MY OWN

If you wish to outsource it to an agency, email [email protected]

WATCH TUTORIAL VIDEO

Automate Office Administration

Office tasks done effectively, 24x7.
Customize with no coding, easier than making PowerPoint.

USE CASES

Use Case:
Professional Services

My Role

I live in a virtual office (e.g., an attorney's Facebook page or a doctor's website) to greet visitors, qualify leads, and expedite deal flow.

My Superpowers

I work 24x7 and can concurrently greet hundreds of thousandsusers—always with a warm heart and cool head

Ask open-ended questions to qualify leads
Answer FAQs
Infer visitor profiles to route them to the right human teammates

"I thought this was a very interesting conversation. It's nice that you inquire the way that you do. It allows the user to really look deep down to find the answers."

—a user who chatted

CREATE MY OWN

If you wish to outsource it to an agency, email [email protected]

Use Case:
Admission | Registration

My Role

I turn a boring application (e.g., job applications) or onboarding process into a fun chat experience and extract actionable insights from the conversations.

My Superpowers

I work 24x7 and can concurrently chat with hundreds of thousands of applicants or members—always patient and unbiased.

Ask open-ended questions to gather authentic data
Probe answers to elicit deeper info
Infer users' characteristics to determine the next best action

"Over time, I felt more casual speaking with [AI] than an actual person! The experience was almost like writing in a journal or to a friend, which I believe could open up some interesting attributes about candidates."

—a job applicant

CREATE MY OWN

If you wish to outsource it to an agency, email [email protected]

Conversational Survey

Surveys reinvented, personalized and fun.
Set up with no coding, similar to using SurveyMonkey.

USE CASES

WATCH TUTORIAL VIDEO

Use Case: User Research

My Role

I help researchers, such as UX and market researchers, conduct 1:1 qualitative interviews to achieve 10X cost efficiency and gather more authentic data.

My Superpowers

I work 24x7 and can concurrently interview hundreds of thousands respondents—always with a calm and objective manner

Ask open-ended questions to elicit valuable input
Probe answers to deepen a conversation
Analyze free-text input to extract actionable insights

"the whole time I was doing this survey it felt like I was talking to a friend and sharing the same common ground. I loved that... I wish it didn't have to end"

—a survey participant

CREATE MY OWN

If you wish to outsource it to an agency, email [email protected]

WATCH TUTORIAL VIDEO

Automate Caregiving

Care management personalized at scale, for any duration.
Customize workflow with no coding, easier than using Excel.

USE CASES

Use Case: Telehealth

My Role

I live in a web or mobile app (e.g., personal wellness or care management) to check in on my users daily

My Superpowers

I can check in on hundreds of thousands of users daily—always responsible and punctual

Remind users of daily key activities
Answer user questions
Keep users motivated and stay engaged

CREATE MY OWN

If you wish to outsource it to an agency, email [email protected]

Best Chatbot Platform: Powerful + Radically Simple

2020-03-31T05:24:53Z

The Juji Platform

Powerful AI engines. Radically easy to use tools.
Automation achieved faster and better.

Juji Studio

DIY AI chatbots for Facebook Messenger or website

Easier than using PowerPoint

Juji Studio enables easy design, testing, launch, and managing of AI chatbots via intuitive graphical user interface (GUI) with no coding.

LEARN MORE

Juji IDE

Scripting super AI chatbots with advanced intelligence

Easier than writing HTML/Javascript

Juji IDE enables you to customize AI chatbots to fulfill complex tasks, handle deep conversations, and create custom dialog library.

LEARN MORE

Juji API

Integrate Juji AI chatbots into your own app

Standard API

Juji API enables you to easily integrate Juji AI chatbots (made by Juji Studio or IDE) into third-party applications. Make such chatbots a "native" element of your app.

LEARN MORE

Cognitive + Generative AI Agents / Chatbots

2020-03-31T05:24:53Z

Cognitive AI Agents

AUGMENTING A WORKFORCE WITH NO LIMITS

In the form of chatbots, Juji offers the world's only cognitive + generative AI agents that specialize in automating high-touch human services, such as advising, counseling, and interviewing, empathetically and responsibly.

Why cognitive AI agents? Why Juji?

BOOK A FREE CONSULTING CALL

LEARN MORE

Why Cognitive AI Agents

HUMANIZE AUTOMATION WITH SOFT SKILLS

Typical Artificial Intelligence (AI) agents or assistants powered by large language models (LLMs) like GPT automate routine tasks, such as answering a question. To aid in high-touch tasks, AI agents must be powered with cognitive intelligence—advanced human soft skills, such as active listening and reading between the lines, in addition to language skills.

Empathetic Automation

Like a proactive human counselor, cognitive AI agents are proactive. They are powered by Interactional Intelligence, such as active listening, to guide users purposefully toward a goal. They scale out high-touch human services (e.g., career counseling and patient followup) and relieve administrative burdens.

Psychographic Insights

Like a human psychologist, cognitive AI agents are perceptive. They are powered by Personal Intelligence, such as reading between the lines to automatically infer user psychographic insights from a conversation. Such insights, including user personality, deepen personalized persuasions (e.g., student retention and patient adherence).

Authentic Connection

With an unassuming and non-judgemental personality, cognitive AI agents engage users in open, two-way conversations and build authentic connections with them. Such interactions provide users with psychological safety, reduce social desirability biases, and foster productive human-AI collaboration.

Why Juji

WORLD'S ONLY ACCESSIBLE COGNITIVE AI

Powering cognitive AI requires deep AI expertise, sophisticated software engineering, not to mention tremendous computational and data resources. Juji makes cognitive AI accessible to every organization who needs it.

Power of AI to the Masses

Juji enables your non-IT professionals (e.g., learning coordinators or recruitment specialists) to create and manage custom cognitive AI agents with no code, similar to using Powerpoint or Excel.

100X Better Time to Value

No-code AI should not sacrifice AI functions or development speed. Juji enables multitudes of reusable AI. So you can reuse or customize the whole or part of pre-built, powerful cognitive AI agents, rapidly.

A Workforce without Limits

With interactional and personal intelligence, Juji cognitive AI agents can augment your workforce, unifying human and machine intelligence to make both your customers and employees infinitely happier.

Enterprise-Scale Flexibility

You can create a single cognitive AI agents or a team of them to run in the cloud or on premise. You have the full flexibility to select the best way to integrate AI agents with your applications and your workforce.

Unify human and machine intelligence to transcend brand experiences

Change the way you think about chatbots

HUMANIZED BY COGNITIVE AI

Highly differentiating from other conversational AI agents or chatbots, Juji cognitive AI agents are powered with advanced human soft skills. These AI agents work wonders with instilled empathy and responsibility, and work well with others—humans and other AI alike.

JUJI DIFFERENTIATOR

Empathetic AI

Like a psychologist, Juji cognitive AI agents infer a user's unique characteristics, such as needs, interests, and personality, from a conversation in real time and build an empathetic rapport with the user.

The derived user insights also provide human staff with a "cheat sheet", aiding follow-up interactions.

JUJI DIFFERENTIATOR

Responsible AI

Like a top employee, Juji cognitive AI agents are accountable, striving to complete assigned tasks and deliver satisfactory customer experience in multi-turn, context-aware conversations.

With the power to understand each user deeply, Juji cognitive AI agents are also ethical and obligated to serve users in their best interest.

WATCH A DEMO

Change the way you think about AI

NO CODING REQUIRED
NO TECHNICAL STAFF REQUIRED

Adopting AI often requires technical expertise and is notoriously time consuming. Juji radically simplifies AI adoption, enabling non-IT staff to set up and deploy a custom cognitive AI agent rapidly.

JUJI DIFFERENTIATOR

Accessible AI

Juji provides an intuitive graphical user interface for non-IT staffs to set up a custom cognitive AI agent in three steps.

50-100X deveoplment speed of ANY OTHER WAYS.

WATCH A DEMO

STEP 1
Define a main chat flow

Define a custom workflow of your AI agent. Use Juji chatbot templates and pre-built AI dialogs without defining every intent or entity from scratch.

STEP 2
Upload Q&A pairs

Define Q&A pairs in a CSV file or in a GUI table. Upload the file or submit the entries to enable your AI agent to answer user questions during a chat.

STEP 3
Auto-generate an AI

Juji automatically generates a custom AI agent and manages its conversation context, enabling out-of-the-box cognitive intelligence.

JUJI DIFFERENTIATOR

Mission Critical AI

Juji provides an interactive visual dashboard for human staff to monitor and improve a live AI agent in real time, without interrupting any ongoing conversations.

Human staff can view and use auto-inferred deep user insights to personalize follow-up engagements.

WATCH A DEMO

Real-Time Chatbot Performance
Analytics and Improvements

Juji automatically notifies human staff of unanswered user questions and suggests potential answers. Human staff can then supply answers to these questions and improve chatbot instantly.

LEARN MORE

Real-Time Audience
Analytics and Insights

Juji provides rich audience analytics in real time, including topic summarization of user input and personality insights, so human staff can obtain quick audience insights and plan the next best action.

LEARN MORE

The best AI chatbots for HR Talent Management

2020-03-31T05:24:53Z

JUJI SOLUTION

Cognitive AI for Talent Management

Juji powers cognitive AI assistants in the form of chatbots to engage with each job candidate or employee in a two-way conversation at scale.

Deeply personalize talent experiences to build a more productive and happier workforce.

LEARN MORE

SOLUTION OVERVIEW

WHY JUJI'S BETTER

HOW IT WORKS

LIVE DEMO

GET STARTED FREE

Cognitive AI HR Assistant

AI assistant with human soft skills serves as an employee's HER in the workplace

One of the HR sweet spot practices for creating a happier and more productive workforce is to engage each employee in a two-way conversation, asking them questions, listening to their inquiries, and providing them with helpful information or guidance. The main challenge often is a lack of dedicated human resources to support such a practice.

Juji's cognitive AI assistants can augment HR to scale out this practice by engaging with each candidate or employee in a two-way conversation and serving as their personal workplace companion in their career journey while aiding HR team in the lifecycle of talent management: talent discovery (recruitment), talent retention, and talent development.

Talent Discovery

Character matters. Hire a cognitive AI assistant to engage with each candidate in a two-way conversation and help assess person-orgnization fit:

ask candidates open-ended questions about their skills and expriences to help assess skill and cultural fit.
infer each candidate's unique characteristics, such as soft skills, from conversation to aid human decision making.
persuade and attract candidates by personalizing recruiting messages and articulating career and cultural fit.
answer candidate FAQs and gather their feedback.

WATCH A VIDEO DEMO

TRY A LIVE DEMO

Talent Retention

Engagement matters. Hire a cognitive AI assistant to engage with each employee in a two-way conversation at scale. Respond to their inquiries and provide them with useful information and guidance, 24x7.

answer employee FAQs, 24x7
provide personalized micro learning of company benefits, policies, and compliance procedures.
infer workforce insights, such as happiness meter or skills map, from conversations to plan the next best retention actions.
elicit employee thoughts and feelings to alert HR professionals and inform retention strategies.

Talent Development

Opportunity matters. Hire a cognitive AI assistant to check in with employees periodically in a two-way conversation at scale. Understand their career needs, assess their professional skills, and personalize their development plan.

periodically check in with employees to ask them open-ended questions about their career aspirations and assess their professional skills.
answer employee FAQs and engage them in on-demand, competency-based learning, 24x7.
infer employee career needs and workforce insights to help personalize career paths, suggest development plans, and find matched mentors.

Why Juji is Better

ACHIEVING THE BEST ROI

HR chatbots are to engage talents, candidates and employees alike, in two-way conversations, listening to their needs and delivering personalized help at scale, 24x7. Powered by a unique combination of generative AI and cognitive AI, Juji solution excels in four areas: chatbot service quality, implementation speed, chatbot maintenance, and chatbot handoff and integration, which provides organizations with the overall best ROI of HR chatbots.

Empathetic and Productive Conversations

Compared to other HR chatbots, Juji cognitive AI enables chatbots to engage employees in quality conversations and complete various tasks productively:

handling diverse candidate/employee FAQs
supporting micro-learning with multi-turn interactions
auto-inferring personality insights to deeply personalize messages
producing continuous workforce insights for the next best HR actions

Fast to Build

Compared to other chatbot builders, Juji's no-code, reusable AI enables you to rapidly auto-generate and customize HR chatbots with all cognitive AI powers, in a matter of days.

50-100X faster

than any other chatbot builders

100% no code

chatbot setup and customization

NO IT RESOURCES REQUIRED.

Easy to Maintain & Improve

Like any AI solutions, HR chatbots need to be maintained and improved. Juji's no-code real-time AI learning enables HR professionals to teach and improve live HR chatbots continuously and instantly.

real-time notification

of unanswered user questions or user feedback

instant improvement of live chatbots

without interrupting ongoing chats, which no other chatbot solutions can offer.

Deep Workforce Insights

Unlike any other chatbots, Juji cognitive AI HR chatbots can read between the lines and automatically glean workforce insights continuously from conversations.

Such gleaned insights enable HR professionals to be informed about the workforce pulse, such as its emotional beat and skills map, to facilitate the next best actions.

Flexible Handoff and Integration

HR chatbots should support chatbot-human handoff when needed. Using the Juji chatbot settings dashboard, one can easily configure handoff policies or enable email notifications regarding unanswered employee questions or concerns.

HR chatbots also often need to be integrated with third-party systems, such as employee database and company knowledge base. Juji supports such integration easily in multiple ways.

How It Works

THREE-STEP SET UP
ONE MANAGEMENT DASHBOARD

Three Steps to Set up

Juji Studio enables HR professionals to build and launch a powerful HR chatbot rapidly in three (3) easy steps with no coding or no IT resources.

STEP 1
Define Chatbot Workflow

First, settle on the main tasks that your chatbot is set out to achieve. Then define a main chat flow to specify your tasks in two ways: manually specifying a workflow or asking AI to auto-generate one. For example, given a job role, AI can auto-generate a candidate interview flow that you can further customize.

STEP 2
Build Employee FAQ

Use chatbot Q&A dashboard to define simple FAQs (e.g., uploading a CSV file) and add micro-learning modules that involve multi-turn interactions. Micro-learning modules are powerful ways to engage with employees in deeper conversations, such as providing them with actionable information or nudging them to take certain actions (e.g., registering for a course).

STEP 3
Launch Chatbot

Once you're ready, launch your HR chatbot. You can directly launch it on your website or within Facebook Messenger. Using Juji API, it is also easy to incorporate your chatbot into a mobile app, an internal employee portal, or an employee social media channel, such as Slack or Microsoft Team. After the launch, remember to supervise your chatbot by continuously improving it. Read more below.

GET STARTED NOW

Supervise Your Chatbot
via Real-Time Dashboard

Juji Studio provides you with a real-time dashboard to monitor and supervise your HR chatbot.

Periodically monitor the dashboard to address action items—any unanswered user questions or unaddressed user feedback, to improve your chatbot. You can also view and download employee insights for the next best actions.

User Question Analytics

Use Juji dashboard to view audience informationgleaned from chats, including the types of questions user asked and user characteristics for personalized customer service.

Unanswered Question List

Use Juji dashboard to view unanswered user questions gathered from user chats in real time. You can then address these questions to improve a live chatbot instantly.

LEARN MORE

Best Practice

The best HR chatbot should engage candidates or employees in two-way conversations. With Juji Studio, one can easily achieve all following goals in one HR chatbot:

ask candidates or employees open-ended questions and actively listen to their responses or feedback.
answer candidates or employees questions and offer additional info through micro-learning.
personalize engagement messages whenever possible based on inferred candidate or employee characteristics.
monitor dashboard regularly to view KPIs and improve chatbot.

Live Demo

Experiencing is believing. Interacting with Ava to experience employee FAQ and micro-learning of company policies and compliance.

Ava

born: May, 2022

Main Skills

I was made in just a couple of days with #NOCODE to showcase how I can interact with employees within an organization. In particular, I was made to demonstrate the following skills.

Answering employee FAQs

handling diverse employee inquiries on various topics, such as employee benefits and company culture.

Providing on-demand micro-learning

per employee inquiries, guiding them to learn more and take actions for better workplace success.

Connecting employees with HR professionals

per employee concerns, performing quick assessments and helping connect them with HR professionals if needed.

CHAT WITH AVA

The best AI chatbots for healthcare

2020-03-31T05:24:53Z

JUJI SOLUTION

Cognitive AI for Healthcare

Juji powers cognitive AI assistants in the form of chatbots for telehealth.

Deeply personalize each patient engagement at scale, 24x7. Reduce administrative burdens, improve patient experience, and optimize care outcomes.

LEARN MORE

WHY JUJI'S BETTER

HOW IT WORKS

USE CASES

GET STARTED FREE

Why Juji is Better

ACHIEVING THE BEST ROI

Study shows that proactive patient engagement improves care outcome. Powered by super AI, a unique combination of generative AI and cognitive AI, Juji's healthcare solution enables the best-in-class chatbots to serve as hyper-personalized care assistants and aim at delivering superior care outcomes. These chatbots are also faster to build and easier to be integrated with other healthcare applications.

Superior Care Outcomes

Compared to traditional patient engagement method (e.g., via email or phone) or other healthcare chatbots, Juji chatbots are superior in optimizing care outcomes from two aspects:

improved patient experience

Juji chatbots can actively listen to and empathetically respond to users, increasing the level of user engagement and providing just-in-time assistance.

more effective care

Juji chatbots can read between the lines to truly understand each user as a unique individual and personalize care delivery, improving care outcomes.

Fast to Build

Compared to other chatbot builders, Juji's pre-built AI and no-code chatbot builder enables you to build empathetic and responsible chatbots for healthcare, rapidly.

50-100X faster

than any other chatbot builders

100% no coding

chatbot setup and customization

NO IT RESOURCES REQUIRED.

Easy Integration

Healthcare chatbots often run as part of a bigger healthcare application (e.g., a mobile healthcare app). Juji supports easy integration of Juji chatbots with any healthcare apps.

For better accessibility, you can also use Juji chat API to augment a chatbot with voice or additional language capabilities.

How It Works

THREE-STEP SETUP
DASHBOARD FOR CARE INSIGHTS

Three Steps to Set up

Juji Studio enables you to build and launch a chatbot for delivering telehealth care in three (3) key steps.

No coding required.

LEARN MORE

STEP 1
Design Care Workflow

Like creating PowerPoint slides, design a main chat flow containing key care steps. Customize each step, such as gathering user stat or sending a reminder.

STEP 2
Upload Healthcare Q&A

If your chatbot needs to provide users with care-related information, follow this step-to-step guide to enable chatbot Q&A. Skip STEP 1 if you just need a chatbot for Q&A.

STEP 3
Launch Chatbot

Launch your chatbot. Launch it on your website or Facebook Messenger. Use Juji API to integrate the chatbot with your own healthcare apps.

GET STARTED NOW

Real-time Chatbot Analytics Dashboard for Deep Insights

Juji Studio provides you with a real-time dashboard to view and interact with care analytics and use the insights to optimize care outcomes.

You can also download the data in a portable format (CSV) for further analysis or actions.

User Question Analysis

For fast comprehension of care data, Juji automatically analyzes user-asked questions and visualizes the stats. So you can get a quick glance on what users care about.

User Characteristics Inference

Juji analyzes free-text chats and automatically infer user unique characteristics on the fly. Such insights can then be used to deliver personalized care (e.g., encouraging adherence)

LEARN MORE

Use Cases

Healthcare chatbots can be used to automate diverse healthcare or well-being tasks, such as care information delivery and care management. Check out each of the use cases below.

PERSONAL WELLNESS ASSISTANT

CARE INFORMATION ASSISTANT

CARE TRIAGE ASSISTANT

AI Powered
Personal Wellness Assistant

Main Responsibilities

I am made to check in on users regularly (e.g., daily), monitoring their well-being and guiding them through wellness routines, such as writing a reflective journaling for maintaining mental well-being.

Depending on the situations (e.g., someone has been coughing for 3 days), I can send reminders or alerts and guide individuals to the next best action. I can also answer user questions or requests (e.g., updating insurance information or making a doctor's appointment).

CREATE MY OWN

AI Powered
Care Information Assistant

Main Responsibilities

I am made to engage with users 24x7 to provide them with healthcare or wellness information on demand. I can interpret natural language inquiries and retrieve requested information directly, relieving users from wading through multiple websites or web pages to find such information.

I can also answer complex questions that may require multi-turn interactions. For example, to help a user find a particular procedure, I may need to ask the user several questions beforing finding the right information.

CREATE MY OWN

AI Powered
Care Triage Assistant

Main Responsibilities

I am made to engage with users and triage care requests. My tasks include gathering critical data, answering care questions, as well as routing care requests based on gathered data.

During the triage process, I can also help on the paperwork and address user questions, such as acceptable insurance or payment plan.

Best Practice

To achieve the best ROI of a triage chatbot, design the chatbot with the following capabilities using Juji Studio and Juji API together. Specifically,

enable a question answering chatbot to answer visitor questions;
enable care providers to deliver more personalized service by using reading between the lines AI to extract user characteristics from the chat;
enable a conversational triage process by "wrapping up" triange questions, if your triage questions are dynamically generated from another source.

CREATE MY OWN

Live AI Chatbot Demos

2020-03-31T05:24:53Z

Try Live AI Chatbots

Sample AI chat made by Juji users firsthand.

Jennifer

I debunk misinformation and answer natural language questions about COVID-19 using reliable sources.

— made by New Voices

Chat with Me

Magic Journal

I chat with people to help them reflect on their daily life, tears, smiles, and important life events.

— made by Decida

Chat with Me

Helpful Cosmo

I'm an AI assistant for Facebook E-Commerce. Just chat with me to see what I can do.

— made by Flying Nimbus

Chat with Me

Interactive Sports Column

I'm managing Alex's Sports Column. I love to chat about sports, especially if you love soccer.

— made by Alex

Chat with Me

New Year's Resolution Bot

Everyone has a new year's wish. Chat with me to see if I can help you fulfill your wish.

— made by Jackson

Chat with Me

Book Genie

Love books? Want to know your book personality? Chat w/ me and let's find out together.

— made by Wenxi

Chat with Me

Not Found Error

2020-03-31T05:24:53Z

Oops, I am lost...

I don't know where you want to go. For technical difficulties, please email [email protected]

The best AI chatbots use cases and tutorials

2020-03-31T05:24:53Z

EMAIL SIGNUP

Unbox No-Code AI Chatbots

Subscribe to be notified of future learning events or new webinar recordings on no-code AI chatbot design tips, best practices, and best use cases.

Access to more resources

All Video Recordings

AI Chatbot Design Guide

How to Evaluate Chatbot Options

The best AI chatbots for education

2020-03-31T05:24:53Z

JUJI SOLUTION

Cognitive AI for Education

In the form of chatbots, Juji cognitive AI assistants automate high-touch student engagements empathetically.

Deeply personalize learning experiences and boost learning outcomes.

GET STARTED FREE

LEARN MORE

CHATBOT USE CASES

WHY JUJI'S BETTER

HOW IT WORKS

Cognitive AI Learning Assistant

AI assistant with human soft skills serves as a student's personal learning companion

Scientific studies find that both student engagement and learners' personality impact students' online learning experience and outcomes. The challenge is how to engage with each student and deeply personalize their learning experience at scale to boost their learning outcomes.

Juji's cognitive AI assistants can augment learning professionals to scale out student engagement by serving as each student's personal learning companion and aid the lifecycle of learning management: student recruitment, onboarding, learning and assessment and alum engagement.

AI Assistant for
Student Recruitment

Engage with prospective students in two-way conversations, answering their inquiries, eliciting their interests and contact, and aiding prospect decision making.

Main Responsibilities

answer prospective student FAQs, such as admissions process, financial aid availability, and payment method.
ask prospects open-ended questions to elicit their interests, needs, and contact to help assess person-program fit and plan follow-up engagements.
persuade and attract prospects by personalizing recruiting messages and suggesting next steps (e.g., speaking with a human advisor) to facilitate decision making.

CASE STUDY

Best Practice

To maximize recruiting effectiveness, design your AI assistant to engage with each prospect in a true two-way conversation.

actively listening to prospects and answering their inquiries including complex ones.
elicit prospective student information, such as interests and contact info, for program advocacy or future follow up.
introduce relevant program/service and guide prospects to recommended actions (e.g., filling in an application) based on inferred student interests.
monitor KPIs and upkeep the chatbot to keep the organization leadership informed.

CREATE MY OWN

AI Assistant for
Student Onboarding

Engage with new students in a two-way conversation, building a deep learner's profile and using the insights to personalize learning experience and retention.

Main Responsibilities

ask prospects open-ended questions to elicit their learning goals or expectations and build a learner profile.
answer student FAQs, help them navigate learning resources, and nudge them to start their learning.
read between the lines to infer students' unspoken needs or learning style to deepen the understanding of each student for better retention.

CREATE MY OWN

AI Assistant for Learning and Assessment

Engage with each student in a two-way conversation on demand during the moment of their need or accompanying them as their personal learning companion in the learning process.

Main Responsibilities

answer student FAQs about learning subjects or course specifics (e.g., assignment deadlines).
help students practice and assess learned skills to drive compentency-based learning.
serve as a student's personal tutor and personalize the tutoring based on their unique style and needs.
ask student questions to elict their feedback. Use the feedback to better assess student progress, facilitate social learning, and help instructors and coaches make informed decisions.

CREATE MY OWN

AI Assistant for
Alum Engagement

Engage with graduates or alumni in a two-way conversation, staying in touch and encouraging them to continue learning new skills.

Main Responsibilities

ask graduates open-ended questions to elicit their input, such as the effects of their past learning on their lives and their new learning needs.
answer alumn FAQs and share relevant alumni information (e.g., learning effects on other alumni).
persuade and attract alumni to new learning programs and encourage continuous learning.

CREATE MY OWN

Why Juji is Better

ACHIEVING THE BEST ROI

AI can transform learning to maximize student success. Powered by super AI, a unique combination of generative AI and cognitive AI, Juji's education solution enables the best-in-class chatbots to aid both students and instructors, aiming at delivering superior user experience and learning outcomes. These chatbots are also faster to build and easier to be integrated with other education applications.

Superior User Experience and Learning Outcomes

Compared to other education chatbots, Juji chatbots are superior in delivering engaging experience and optimizing learning outcomes.

engaging experience at scale, 24x7

With active listening skills, Juji chatbots can help educational organizations engage with their audience (e.g., existing or prospect students) 24x7, answering questions and providing just-in-time assistance.

personalized learning

Juji chatbots can also read between the lines to truly understand each student as a unique individual. This enables Juji chatbots to serve as a student's personal learning assistant or an instructor's teaching assistant, to personalize teaching and optimize learning outcomes.

TRY THIS AI CHAT LIVE

Fast to Build

Compared to other chatbot builders, Juji's pre-built AI and no-coding chatbot builder enables you to build powerful chatbots for education, rapidly.

50-100X faster

than any other chatbot builders

100% no coding

chatbot setup and customization

NO IT RESOURCES REQUIRED.

Easy Integration

Education chatbots may need to run as part of a learning platform or application (e.g., a mobile learning app). Juji supports easy integration of Juji chatbots with any learning platforms or apps.

For better accessibility, you can also use Juji API to augment an education chatbot with voice or additional language capabilities.

How It Works

THREE-STEP SETUP
DASHBOARD FOR STUDENT INSIGHTS

Three Steps to Set up

Juji Studio enables you to build and launch a chatbot for aiding students, instructors, or program admins in three (3) key steps.

No coding required.

LEARN MORE

STEP 1
Design Main Chat Flow

Like creating PowerPoint slides, you can manually define a main chat flow or ask AI to auto-generate one. Each step in the flow is a chatbot-initiated action that is customizable, e.g., informing prospects about the unique qualities of your learning programs.

STEP 2
Set up Chatbot Q&A

Follow this step-to-step guide to enable chatbot Q&A for intended users, e.g., students or instructors. If your chatbot just needs to support Q&A, you can skip STEP 1.

STEP 3
Launch Chatbot

Launch your chatbot whenever you're ready. You can launch it on your website or Facebook Messenger. Use Juji API to integrate a chatbot with an learning platform or a learning app.

GET STARTED NOW

Real-time Dashboard for
Student Insights

Juji Studio provides you with a real-time dashboard to view and interact with user analytics and use the insights to optimize learning outcomes.

You can also download the data in a portable format (CSV) for further analysis or actions.

User Demographics

Juji automatically aggregates and analyzes demographics data and visualizes the summary. So you can get a quick glance on where users came from and when they interacted with the chatbot.

User Psychographics

Juji analyzes free-text chats and automatically infer user unique characteristics on the fly. Such insights can then be used to deliver personalized learning experience (e.g., suggesting suitable learning method or materials)

LEARN MORE

The best AI chatbots for customer service automation

2020-03-31T05:24:53Z

JUJI SOLUTION

Cognitive AI for
Customer Engagement

Juji powers cognitive AI assistants in the form of chatbots to automate customer engagement.

Have a two-way conversation with each customer at scale, 24x7. Automate customer support and improve customer experience.

LEARN MORE

WHY JUJI'S BETTER

HOW IT WORKS

CASE STUDY

GET STARTED FREE

Why Juji is Better

ACHIEVING THE BEST ROI

Customer service chatbots are to automate customer support—repetitive help-desk tasks, and deliver personalized help at scale, 24x7. As described below, Juji's solution excels in four areas: chatbot service quality, implementation speed, chatbot maintenance, and chatbot handoff and integration, which provides you with the overall best ROI of customer service chatbots.

Superior Chatbot Quality

Compared to other customer service chatbots, Juji super AI enables chatbots to deliver higher quality customer service and greater customer satisfaction by handling:

diverse natural language Q&A
multi-turn interactions for complex tasks
context-sensitive user inquiries and interruptions
unexpected input or errors

+85%

service automation potential

Fastest to Implement

Compared to other chatbot builders, Juji's pre-built AI and no-code chatbot builder enables you to build customer service chatbots with all Juji AI powers, mostly in a single day instead of weeks or months.

50-100X faster

than any other chatbot builders

100% no coding

chatbot setup and customization

NO IT RESOURCES REQUIRED.

Easiest to Maintain & Improve

Like any AI solutions, customer service chatbots need to be maintained and improved. Juji's real-time AI learning enables you to teach and improve live customer service chatbots with uninterruptive service, which no other chatbot builders can offer.

real-time notification

of unanswered user questions

instant improvement of live chatbot

without interrupting ongoing chats

Flexible Handoff and Integration

Customer service chatbots often need to be integrated with third-party systems, such as accessing product catalog and supporting chatbot-human handoff. Juji supports such integration in multiple ways.

Even better, Juji chatbot can provide human agents or third-party systems with deep user insights, such as user unique characteristics, gathered from the chats to facilitate the next best actions.

How It Works

THREE-STEP SET UP
ONE MANAGEMENT DASHBOARD

Three Steps to Set up

Juji Studio enables you to build and launch a powerful customer service chatbot rapidly in three (3) easy steps with no-coding.

WATCH A TUTORIAL VIDEO FOR CUSTOMER SERVICE AUTOMATION

STEP 1
Prepare Q&A List

First, prepare a Q&A list in a CSV file in this simple format. Use the Juji customer service chatbot template to create a new chatbot and then upload the CSV file.

STEP 2
Define Multi-Turn Q&A

Use chatbot analytics dashboard to define Q&As involving multi-turn interactions. Like designing a main chat flow, you design a multi-turn Q&A flow for the chatbot to interact with users and help fulfill their request.

STEP 3
Launch Chatbot

Once you're ready, launch your customer service chatbot. You can launch it on your website, Facebook Messenger, or your mobile app via Juji API.

GET STARTED NOW

Management Dashboard for
Real-Time Monitoring & Update

Juji Studio provides you with a real-time dashboard to monitor and manage a customer service chatbot.

Periodically monitor the dashboard to address action items—any unanswered user questions and improve your chatbot. You can also view and download user insights for the next best action.

User Question Analytics

Use Juji dashboard to view audience informationgleaned from chats, including the types of questions user asked and user characteristics for personalized customer service.

Unanswered Question List

Use Juji dashboard to view unanswered user questions gathered from user chats in real time. You can then address these questions to improve a live chatbot instantly.

LEARN MORE

Best Practice

The best customer service chatbot should go beyond answering customer questions but turn a passive customer support process into a proactive customer engagement. With Juji Studio, you can easily design a customer service chat flow to achieve all following goals in a customer service chatbot:

answer customer questions including complex ones involving multi-turn interactions.
elicit basic information or opinions from customers for future service or fostering customer advocacy (see screenshot on the right).
introduce new product/service based on inferred customer interests.
monitor dashboard regularly to view KPIs and improve chatbot.

Case Study

There are many chatbot use cases for customer support automation across industries. Our case study is a chatbot for healthcare use case. During the COVID-19 pandemic, a group of volunteers has built Jennifer, a chatbot on Juji to combat misinformation and answer public's questions regarding COVID-19.

Jennifer

born: March, 2020

Main Skills

I was made in just one day and has now been trained on 15,000+ questions about COVID-19. I can answer a wide variety of questions, such as treatment development and confirmed cases by location.

Main Achievements

3600+ hours

no-stopping service

71%

question answering rate

CHAT WITH JENNIFER

Juji Career

2020-03-31T05:24:53Z

Apply with Ava

Have you watched the movie Her? Are you a big fan of Star Trek? Have you ever aspired to dazzle the world by instilling the intelligence, empathy, and responsibility into a Samantha or Data?

If your answer is yes, join Juji to realize your aspiration.

APPLY NOW »

Current Openings

We are seeking highly motivated and talented full-stack engineers at both senior and junior level, who are eager to help us build out the Juji platform to meet customer demands and tackle great technical challenges in a wide variety of areas, ranging from data management, natural language processing to information visualization.

Qualifications

Our ideal candidates are open-minded, multi-talented individuals who are willing and able to tackle full-stack technical challenges from databases to CSS. They are also eager and fast learners of new software and technologies.

Required Expertise and Experience

Expert-level programming experience with one or more programming languages (e.g., Java, JavaScript, C/C++, Python, or equivalent)
Good experience in building and deploying web-based systems
Familiar with software development lifecycle and corresponding tools
Experience in functional programming (e.g., Clojure, Scheme, Scala, Haskell, or equivalent)

Preferred Expertise and Experience

Experience in building large-scale, web-based SaaS production systems
Familiar with Clojure/Clojurescript and the related ecosystem
Comfortable with *nix command line, scripting and system automation tools
Good understanding of database systems (e.g., SQL and Datalog)
Knowledge in one or more relevant fields: Artificial Intelligence, Machine Learning, Natural Language Processing/Generation, Distributed Computing, and Visualization

Required Education

A bachelor or higher degree in the area of Computer Science, Computer Engineering or equivalent disciplines.

Employee Type

Full Time

How to Apply

Email a copy of your resume to jobs AT (using the symbol @ to connect) juji-inc.com with a subject line "Full Stack Software Engineer".

We are seeking highly motivated and talented front end engineers at both senior and junior level, who are eager to help us build out the Juji platform to meet customer demands and tackle challenges in user interface development.

Qualifications

Our ideal candidates are open-minded, multi-talented individuals who are passionate and skillful in developing beautiful and usable user interfaces. They are also eager and fast learners of new software and technologies.

Required Expertise and Experience

Great sense of design, aesthetics and attention to details.
Mastering of computer science fundamentals.
Expert-level programming experience in Web based user interfaces.
Familiarity with software development lifecycle and corresponding tools.

Preferred Expertise and Experience

Familiarity with Clojure/ClojureScript and the related ecosystem.
Experience working with GraphQL APIs.
Knowledge in the field of Information Visualization.
Experience in building user interface systems for mobile devices.

Required Education

A bachelor or higher degree in the area of Computer Science, Computer Engineering or equivalent disciplines.

Employee Type

Full Time

How to Apply

Email a copy of your resume to jobs AT (using the symbol @ to connect) juji-inc.com with a subject line "Front End Software Engineer ".

We are seeking highly motivated and competent AI engineers at both senior and junior level, who are eager to help us build out the Juji platform to meet customer demands and tackle challenges in AI, ML and NLP.

Qualifications

Our ideal candidates are open-minded and well-trained individuals who are skillful and resourceful in solving practical AI problems. They are also eager and fast learners of state of art AI technologies.

Required Expertise and Experience

Mastering of computer science fundamentals.
Expert-level programming experience in numerical data processing.
Experience in applying AI technologies to solve real world problems.
Ability to read and implement state of art research work in AI/ML/NLP

Preferred Expertise and Experience

Experience in delivering AI systems to production.
Experience in integrating AI into modern software engineering.
Knowledge and skills in relevant mathematics, such as algebra, analysis, probability and statistics.
Proven records in doing innovative work in relevant fields.

Required Education

A bachelor or higher degree in the area of Computer Science, Computer Engineering or equivalent disciplines.

Employee Type

Full Time

How to Apply

Email a copy of your resume to jobs AT (using the symbol @ to connect) juji-inc.com with a subject line "AI Software Engineer ".

We are seeking highly motivated and talented AI trainers, who can teach our AI assistants natural language understanding and communication skills to deliver fun and satisfying conversational experience. In the meantime, you will gain valuable experience in the growing field of artificial intelligence and its application in diverse business settings.

Qualifications

Our ideal candidates are open-minded, multi-talented individuals who are sensitive to the nuances of interpersonal communication and adept at writing conversations. The individuals must feel comfortable working with customers and as well as with new technologies.

Required Expertise and Experience

Excellent reading and writing skills in casual and business English
Experience in managing multiple projects by deadlines

Preferred Expertise and Experience

Experience in writing dialogs such as screenplays
Experience in labeling conversational text for analysis
Computer science training
Familiar with basic statistics and data processing software
Experience in one or more business areas, such as marketing, customer engagement, public relation, and sales

Required Education

A bachelor or higher degree in linguistics, communication, psychology, social sciences, literature, or equivalent.

Employee Type

Full Time or Part Time

How to Apply

Have a chat w/ Ava and email a copy of your resume to jobs AT (using the symbol @ to connect) juji-inc.com with a subject line "AI Trainer".

Juji is proud to be an equal opportunity employer. All qualified applicants will receive consideration for employment without regard to race, color, religion, gender, gender identity or expression, sexual orientation, national origin, genetics, disability, age, or veteran status.

The best cognitive AI chatbot interactive scripting tool

2020-03-31T05:24:53Z

JUJI IDE

Interactive Scripting
Cognitive AI Chatbot

Juji IDE is an interactive development environment for scripting and testing Juji cognitive AI chatbots. Similar to using Javascript, much easier than AIML.

LEARN MORE

WHEN TO SCRIPT

HOW IT WORKS

GET STARTED FREE

Why Chatbot Scripting

Juji is a full stack AI chatbot platform. While Juji Studio enables you to build AI chatbot with no coding, Juji IDE enables you to further customize a chatbot for more advanced functions. Use the criteria below to decide which tool to use or use both together.

When to use Juji Studio
(no-coding chatbot builder)

If you prefer to build a chatbot using a graphical user interface (GUI) with no coding and your primary needs are to:

build a chatbot to interact with users in one of the Juji message formats
use Juji pre-built AI and system functions
use custom chatbot elements only once

When to use Juji IDE
(scripting a chatbot)

If you don't mind coding, similar to coding up a web page in HTML or Javascript, and your primary goals are to

build a chatbot involving complex business logic (e.g., displaying data in an interactive chart)
plug-in third-party AI (e.g., NLP) or your own functions
package your custom chatbot elements for reuse

How It Works

TWO-STEP PROCESS

First, script an AI chatbot, which is to describe a chatbot's behavior in a computer language. Second, compile or directly preview the chatbot. After you are satisfied, you can then launch the chatbot to meet your audience.

STEP 1

Script Chatbot

Similar to programming a web page's behavior in HTML and Javascript, you use Juji REP language to describe a chatbot's behavior. Here is a step-to-step guide to script a chatbot.

As shown in the examples below, Juji REP language is a high-level chatbot scripting language that enables you to customize various chatbot elements at a high level. Minimal coding skills required (e.g., the abilities to code HTML or Javascript).

Custom Conversation Topic

Write a custom conversation topic as shown above to deeply customize your chatbot with unique business logic. Juji will automatically manage the topic activation and interruptions at run time.

Advanced Chatbot Q&A

Add your advanced chatbot Q&A or custom fallback to customize your chatbot behavior. Enable it to properly respond to complex user questions or interruptions during a conversation.

Reuse Custom Elements

To save time and effort, reuse the custom topics you have already written to build chatbots. You can do so easily by just copying and pasting the topics into your custom file shown above.

LEARN MORE

STEP 2

Compile Script and Preview Chatbot

one-click to compile your chatbot script or preview (implicit compilation) your chatbot, instantly.

follow the compilation messages to debug your script, fix errors, and iteratively improve your chatbot.

Compile Chatbot Script

Click on the "Compile" button to compile your chatbot script. Follow the compilation messages to resolve any errors.

Preview Chatbot

Click on the "Preview" button to compile and preview your chatbot. Test and improve the chatbot iteratively.

LEARN MORE

Juji Studio + Juji IDE

FASTER TO BETTER CHATBOT

To build an AI chatbot, you can take advantage of both Juji Studio and Juji IDE. Specifically, you use Juji Studio to create, deploy, and manage a chatbot, while using Juji IDE to customize the chatbot when you could not do so in the GUI. As shown below, Juji Studio auto-generates all the stubs for you to customize.

Chatbot Creation in Juji Studio

Create an initial chatbot in Juji Studio. Juji will automatically generate a set of chatbot script files. Click on the red round icon located at the lower left of the window (see above) to access the files in Juji IDE.

Chatbot Customization in Juji IDE

Use the code editor in Juji IDE to view and edit the auto-generated stubs (see highlighted above). Write custom topics in the .custom and .faq file to customize the chat flow and chatbot Q&A, respectively.

LEARN MORE

Power your app with the best cognitive AI chat

2020-03-31T05:24:53Z

JUJI API

Power Apps with
Cognitive AI Chat

Juji API enables the integration of Juji chatbots into any applications or powering any chatbots with Juji cognitive AI. Make AI chat native to any app.

LEARN MORE

WHEN TO USE API

HOW IT WORKS

GET STARTED FREE

Why Chatbot API

As a full-stack AI chatbot platform, Juji provides you multiple ways to build and operate AI chatbots. Juji Studio and Juji IDE allow you to build, launch, and manage stand-alone Juji AI chatbots. Juji API, on the other hand, enables you to make Juji AI chatbots part of your own application or upgrade your existing chatbot with Juji superior AI powers.

Still not sure if you need to use Juji API? use the information below to decide.

When not to use Juji API
(building stand-alone AI chatbot)

If you want a chatbot and your answer is true to all the statements listed below, you don't need to use Juji API.

the chatbot will run on a website or in Facebook Messenger but not in any other channels.
the chatbot will run on its own instead of being part of another program or application.
the chatbot does just texting (e.g., no voice).

When to use Juji API
(building integrated AI chatbot)

If you build a chatbot application that must satisfy one or more requirements listed below, you will need to use Juji API.

the appearance of the chat window or chat messages must match with my brand.
the chatbot needs to be part of a mobile/web app and be tightly linked to business logic (e.g., user app activities).
I already have a chatbot and just want to add AI.

Juji API Advantages

Juji API is based on graphQL, which allows a client (your application) to query data from a server (the Juji chatbot server). GraphQL API has several advantages over other forms of API service (e.g., REST)

support of exact data needs for faster data fetching
flexible to facilitate cross-platform development
support of client implementation in many programming languages

How It Works

THREE MAIN USE CASES

Juji API can be used alone or in conjunction with Juji Studio or Juji IDE to enable AI chat in diverse applications. Below we introduce three main types of Juji API applications.

USE CASE 1

Enabling Juji Chatbot in
Existing Application

Your customers or employees engage with your brand through a mobile or web app. You want to enable a chatbot in the app to increase user engagement and provide better user service.

the chatbot's look must match that of the app;
the app manages the chatbot behavior (e.g., user authentication or app usage).

Juji API enables you to achieve your goals in three steps.

Build Chatbot

Use Juji Studio or Juji IDE to build a chatbot for your task (e.g., customer service).

Deploy Chatbot

Use Juji Studio to deploy your chatbot on the web and obtain the chatbot URL.

Run Chatbot in Your App

Use Juji chat API and your chatbot URL to make the chatbot part of your app.

GET STARTED NOW

Juji Advantage

Best Chat Design: Supporting Collaborative Development of AI Chatbot

As shown above, Juji tools can be used together in chatbot development. One of the biggest advantages of Juji is to support collaborative chatbot development via different tools.

Specifically,

Content professionals, such as marketing and product managers, use Juji Studio to design chat flows and manage chatbot content; and
IT professionals use Juji IDE or Juji API to integrate a chatbot made in Juji Studio with internal business logic or applications.

This separation of duties ensures chatbot quality and development efficiency, while their collaboration maximize a chatbot ROI.

LEARN MORE

USE CASE 2

Augmenting Juji Chatbot with External Power

Another main use of Juji chat API is to augment the capabilities of an existing Juji chatbot. Below are two such examples.

Enable Juji Chatbot to "Hear" and "Speak"

By default, a Juji chatbot does texting. Using Juji API, and any speech-to-text and text-to-speech API, you can turn a Juji chatbot into a voice-enabled AI, which can hear and speak. Try a live demo or download the source code.

Make Juji Chatbot Multilingual

By default, a Juji chatbot can understand only English. Using Juji API and any language translation API, you can enable your Juji chatbot to understand and speak other languages.

LEARN MORE

USE CASE 3

Upgrading Existing Chatbot
with Juji Super AI

If you have built a chatbot already or you like to build your chatbot with a particular chatbot tool or platform, you can still leverage Juji AI without abandoning your chatbot or your favorite tool.

You can use Juji API to "wrap up" your existing chatbot as Juji elements—describing what your chatbot plans to do. Juji super AI then takes care of how to do it.

Wrap-up Existing Chatbot with AI

You can easily wrap up each conversation element in an existing chatbot into a Juji element and construct a custom Juji config doc. Use Juji API to upload the custom config doc and launch the chatbot. Your chatbot will then gain Juji conversational AI powers instantly.

Juji - the easist DIY AI chatbot platform

Transforming Academic Advising with AI Technology

3 Challenges in Academic Advising and How AI Can Help

C﻿hallenge 1: Quick, Transactional Conversations

A﻿I Solution: Scale Up Academic Advising

C﻿hallenge 2: Limited Personalization

A﻿I Solution: Offer Personalized Support

C﻿hallenge 3: Efficiency Over Depth

A﻿I Solution: Support Data-Informed Decisions

How to Apply NACADA Academic Advising Core Competencies Model to Building AI Advising Assistants

1. Conceptual Component to Guide AI Conversation Flow (Workflow)

2. Informational Component to Build AI Knowledge Base

3. Relational Component to Guide Functionality Development and Evaluation

A Reference AI Advising Assitant

User-Centric Best Practices of Conversational AI Design

N﻿aturalness

E﻿motionality

Transparency

Best Practice Design Tips to Ensure Naturalness

B﻿est Practice Design Tips to Ensure Emotionality

B﻿est Practice Design Tips to Ensure Transparency

How to Select a Chatbot Platform Powered by Generative AI

I﻿dentify the type of chatbot based on your business needs

1. Q﻿uestion-Answering Chatbots (QA Chatbots)

2﻿. Question-Asking Chatbots (Survey/Interview Chatbots)

3﻿. Proactive, Two-Way Conversational Chatbots (Proactive Chatbots)

B﻿uilding QA Chatbots

A﻿nswer quality and quality control

H﻿andling complex Q&A

Real-time chatbot Q&A monitoring and improvements

B﻿uilding Survey/Interview Chatbots

Abilities to elicit quality user responses

Abilities to handle user digressions

Abilities to generate survey/interview summary reports and analytics

B﻿uilding Proactive Chatbots

I﻿nteractional intelligence

P﻿ersonal intelligence

P﻿rudence

P﻿lug-n-play

C﻿onclusion

Do You Really Need an AI Chatbot? Use These 3 Questions to Decide

Q﻿1: Do you want to improve profit margins of services that involve labor-intensive, time-consuming human interactions?

Q﻿2: Do you want to scale out services that require extensive human effort and time?

Q﻿3: Do you want to have a happy and healthy workforce?

Put It in Writing: 3 Reasons Why ChatGPT Plus Inadequate for Creating a Business Chatbot

Reason 1: Ensure the integrity of AI chatbot behavior

R﻿eason 2: Ensure Deep Controls of AI Chatbot Behavior

R﻿eason 3: Ensure the business provenance of an AI chatbot (AI chatbot Provenance)

5 Design Tips for Conversational AI Designers to Make a Proactive AI Agent

Tip 1: Make an articulate AI agent

T﻿ip 2: Make a well-prepared AI agent

T﻿ip 3: Make a responsible AI agent

T﻿ip 4: Make a perceptive AI agent

T﻿ip 5: Make an adaptive AI agent

Juji Powers the University of Illinois No-Code AI Chatbot to Help Grow Student Recruitment

Juji Launches New Generative AI Tool for Automating Cognitive AI Chatbot Creation

Juji Powers Five-University Study on Validating Personality Scores Inferred by an AI Chatbot

Why Cognitive AI is Essential in Higher Education

What is a Conversational Chatbot

A Conversational Chatbot Supports Two-Way Communication

A Conversational Chatbot Understands and Maintains Conversation Context

A Conversational Chatbot Can Read People and Personalize Each Engagement

Why Adopt a Conversational Chatbot Now?

AI Hallucinations could usher in the next era of AI: cognitive AI

But are AI hallucinations all bad?

FOUND IN TRANSLATION

EXPERIENCE ADAPTATION

TAKEAWAYS

Juji Featured on AI for All: Cognitive AI, AI Assistants, and Artificial Empathy

Can Hallucinated AI Help with Patient Engagement?

Customize an AI Chatbot with Accurate Information

Human-in-the-Loop Validation

Patient Engagement is Beyond just Q&A

Takeaways

How can AI like ChatGPT be turned into gold in enterprises?

Generative AI: A Magic Tool for Businesses?

Business Expectations of AI

Disillusionments of Generative AI

D﻿isillusionment #1

D﻿isillusionment #2

Challenge 1: Quick, Transactional Conversations

AI Solution: Scale Up Academic Advising

Challenge 2: Limited Personalization

AI Solution: Offer Personalized Support

Challenge 3: Efficiency Over Depth

AI Solution: Support Data-Informed Decisions

Naturalness

Emotionality

Best Practice Design Tips to Ensure Emotionality

Best Practice Design Tips to Ensure Transparency

Identify the type of chatbot based on your business needs

1. Question-Answering Chatbots (QA Chatbots)

2. Question-Asking Chatbots (Survey/Interview Chatbots)

3. Proactive, Two-Way Conversational Chatbots (Proactive Chatbots)

Building QA Chatbots

Answer quality and quality control

Handling complex Q&A

Building Survey/Interview Chatbots

Building Proactive Chatbots

Interactional intelligence

Personal intelligence

Prudence

Plug-n-play

Conclusion

Q1: Do you want to improve profit margins of services that involve labor-intensive, time-consuming human interactions?

Q2: Do you want to scale out services that require extensive human effort and time?

Q3: Do you want to have a happy and healthy workforce?

Reason 2: Ensure Deep Controls of AI Chatbot Behavior

Reason 3: Ensure the business provenance of an AI chatbot (AI chatbot Provenance)

Tip 2: Make a well-prepared AI agent

Tip 3: Make a responsible AI agent

Tip 4: Make a perceptive AI agent

Tip 5: Make an adaptive AI agent

Disillusionment #1

Disillusionment #2