What Can I Do With MARANICS?

Every now and then, we get to witness a moment that perfectly captures why MARANICS exists. Not a marketing line nor a feature list, just a real moment, one that happens when people are empowered to shape the tools they use every day. A moment that we are proud of.
Recently, one of our customers shared a story that does exactly that.

A New Use Case… Hidden in Plain Sight
This team had just started using MARANICS. They were still exploring, still learning, still discovering what was possible. One day, a regular user in the company decided to build a new use case directly in the system. They didn’t even announce it. It simply went live.

Less than 24 hours later, another employee stumbled upon it while starting a different task. It caught their eye, but since it wasn’t immediately relevant, they continued with their work. And then something fascinating happened.

Discovery Without Training
Later during the same shift, that new use case became applicable, and the user remembered seeing it. No onboarding session. No walkthrough. No training videos. They simply recognized it and used it. And not only did it work… It worked brilliantly.
The employee completed the process faster than ever before. The supervisor was amazed, not just by the speed, but by the fact that the user felt confident navigating a brand-new workflow independently. When systems are intuitive, people don’t need instructions, they just start using them.

The Best Part? It Wasn’t Built by Us

The use case wasn’t built by MARANICS. It was built by a regular user that had a problem they needed to solve.
This is our mission in action. MARANICS isn’t here to lock organizations into rigid systems that only designers or developers can build in. We’re here to make operational design accessible to the people closest to the work, the ones who understand real-world needs better than anyone. When users can build, publish, and improve their own workflows, something powerful happens: your organization starts evolving from the inside out.

A Glimpse of What’s Possible
Stories like this remind us that digital transformation doesn’t have to be massive, complex, or top-down. Sometimes, it’s a single employee creating a single use case that unlocks a new way of working for everyone else. When that happens organically, you know you’ve built something that truly works for people.

Why This Matters
This story highlights a few things we believe in:
Intuitive systems reduce training overhead.
If a user can adopt a new workflow instantly, friction disappears.
Empowerment drives innovation.
You don’t need a project. You need a moment or an idea, and a system that lets you act on it, like Maranics.

Scaling Use and Operational Impact

As the fleet expands, usage continues to rise. Crews across all vessels now complete over 4,500 digital checklists each month, including many newly deployed installations still in their first operational phase.
Our most active vessels, together with their crews, complete more than 800 checklists per month – roughly one every hour. Across the rest of the fleet, we see steady growth as Maranics becomes an integral part of daily operations.

Each completed checklist contributes structured, high-quality data that combines human observations, system inputs, and sensor readings into one contextual record. The result is a live operational picture that helps crews and shoreside teams make faster, better-informed decisions while maintaining full compliance.

Why Vessels Choose Maranics

Vessels use Maranics to connect human activity with machine and sensor data, ensuring that every event at sea or in port carries the full story behind it. Routine tasks such as inspections, maintenance, and navigation checks become continuous data flows that drive insight and accountability across the operation.
This combination of human context and digital precision transforms daily workflows into an ongoing source of learning. It supports predictive maintenance, improves safety management, and builds a reliable data foundation for future AI-driven performance analysis.

Global Reach

The heatmap shows where Maranics-powered vessels are active today. Our footprint now spans North America, Europe, and Asia, with growing presence along the world’s busiest maritime routes. Each new deployment extends that network, strengthening the shared flow of operational knowledge that connects ship and shore.

Thanks to our partners and team for making this possible!

Maranics provides the missing link. By combining machine data with human input, we create context-rich datasets that AI agents and teams across domains can actually understand and use.

Why Context Matters

Machine data captures what happened. Human input explains why it happened, how it was handled, and under what conditions. Without this layer:

• AI agents misinterpret anomalies.
• Teams across functions lack a shared understanding.
• Critical decisions are made without traceability or rationale.

As we noted in our World Maritime Day reflections, “up to 95% of the most important operational data can only be captured by humans.”

Industry research reinforces this point. Human-in-the-loop data pipelines preserve context and accountability at points where automation alone falls short.

[..] data alone doesn’t yield valuable information. It takes people to analyze and interpret that data for insights – The Human Factor

Checklists as a Core Mechanism

One of the most effective ways we capture human data is through our checklist app.

• Template-driven: checklists can be defined once and instantly reused.
• Simple interface: easy for crews and operators to adopt.
• Instant updates: when processes change, checklists adapt immediately without code deployments.
• Structured by design: every entry is validated, timestamped, and enriched with metadata.

This flexibility makes Maranics a tool that can react instantly to process changes. Evidence from healthcare and aviation shows the same principle: digital checklists increase compliance and improve performance in high-risk settings (See: Comparing the Effects of Paper and Digital Checklists).

From Data to Understanding

By combining machine signals with human input, Maranics transforms fragmented data into a complete picture:

1. Capture – systems log events, humans add context via checklists.
2. Enrich – inputs are validated and tagged with metadata.
3. Correlate – machine and human data are linked together.
4. Trigger – contextualized data drives workflows and alerts.
5. Interpret – AI and teams see meaning, not just numbers.
6. Adapt – processes evolve instantly through updated templates.

The result: data that is not just collected but understood.

Use Cases

• Operations – reduce false alarms by pairing system alerts with human confirmation.
• Compliance – build auditable records enriched with human rationale.
• Cross-domain collaboration – share contextualized data across engineering, operations, and compliance.
• AI enablement – train and run AI agents on data that contains context, reducing false positives. Industry experience shows that human investigators are essential to avoid false positives in sensitive domains such as fraud detection (See: The Role of Human-in-the-Loop). Avoiding the “crap in, crap out” problem ensures AI delivers reliable results (See: Data Quality in AI).

See It in Action

Want to see how this works in practice? Our video library shows the platform in real-world scenarios, capturing human input, enriching machine data, and driving processes forward.

Looking Ahead

We continue to extend the platform with new capabilities:

• Hands-free interaction – enabling checklist use in constrained environments.
• Progress indicators – visualizing the status of multi-step operations.
• Image recognition – reading expiry dates on fire extinguishers, safety signs, and labels to automatically capture and contextualize information.
• Data usage visibility – showing where and how data points are reused across workflows.

Conclusion

Maranics bridges the gap between raw data and real understanding. By embedding human context into machine data through checklists, workflows, and automation, we turn isolated inputs into meaningful, actionable information.

Without context, data risks falling into the “crap in, crap out” trap. With Maranics, your data doesn’t just exist – it makes sense, drives better decisions, and powers AI that you can trust.

more than 80 vessels are now running Maranics.

And the pipeline is only getting stronger. Some major industry names are preparing to come onboard, which will eventually double that number. But vessel rollouts at large enterprises don’t always move quickly — the unpredictable course of today’s political and economic climate makes planning ahead more complex than ever.

From Growth to Flexibility

That’s exactly why deployment flexibility matters. When the future is uncertain, customers need confidence that they can adopt Maranics in a way that fits their current and future reality. To support this, we provide multiple deployment paths:

• Hosted Cloud (Zero Setup, Secure by Design)
  Every customer gets access to our hosted backend with database-level separation for maximum data security. No setup required — just log in and go.
• High Seas Deployment (Offline Ready)
  For vessels without a stable connection, we can deploy a local server directly on the ship. It runs fully offline, and once a connection is available, it automatically syncs with the cloud. The crew doesn’t need to lift a finger — reports flow seamlessly.
• Self Deployment (Full Control)
  Some customers prefer everything in-house. For them, we offer a self-deployment option to run our cloud endpoint on-premise, keeping operations fully under their control while meeting local requirements.

Heavy Usage, Clear Signal

And not all challenges are setbacks. The vessels already running Maranics are using it heavily – some generating 800+ checklists every month, more than one checklist every single hour. That level of adoption is a clear signal: once onboard, Maranics becomes an essential part of daily operations.

We’re excited about the growth, the upcoming large-scale rollouts, and the ways customers are already relying on Maranics every day. Whether in the cloud, at sea, or fully in-house — we’re building a platform that adapts to our customers’ world and keeps getting stronger.

Digital checklists are already a great fit for many onboard tasks. But there are moments, such as when wearing gloves, handling equipment, or operating in tight spaces, where even the best interface takes a back seat to simply speaking. In those cases, having an assistant that listens and responds can make routine procedures faster, smoother, and more natural to complete.

From Mockup to Interaction

The idea is simple: speak naturally and let the assistant fill in the checklist for you. Our prototype allows testers to say things like “Start departure checklist,” confirm with “Yes, all mooring lines removed,” or log values like “Main engine temperature 410 degrees Celsius.” The assistant interprets each response, matches it to the correct field, and updates the checklist.

This isn’t just voice-to-text. The assistant uses the same logic structure that powers our metadata-rich checklists. It knows which fields are required, adapts based on your answers, and keeps track of what’s already been said.

We’re currently evaluating clarity of prompts, pacing of the conversation, and how it feels to be guided through a workflow by voice. Feedback from early testers has helped us refine the interaction to feel helpful, not robotic.

Testing the Real Challenge

One of the most important lessons so far isn’t about AI accuracy—it’s about deployment. Vessels often operate without internet. For this assistant to be useful, it must run offline. That means deploying it on local edge devices that can handle the compute load without disrupting other onboard systems.

We’re working to ensure the assistant runs smoothly alongside other tools, without slowing down the system or draining shared resources. In the future, we’d love to run the assistant directly on mobile phones. But that depends on hardware availability, and most current devices in the field aren’t yet up to the task.

What’s Next

The voice assistant is still in evaluation, but we’re making steady progress. We’ll continue tuning the interaction, exploring noise resilience, and refining how the assistant performs in resource-constrained environments.

If you’re curious about how this could work for your crew or want to be part of the early feedback loop, get in touch.

Why Reranking Matters

Vector search helps us find data based on meaning by comparing the similarity between a user’s query and stored entries. This is great for capturing related information, but it often returns results that are similar but not always aligned with what the user actually wants. That’s where reranking comes in.

A reranker looks at the context of the query and the returned results to understand the user’s true intent. It then reorders the similar entries from the vector search, ranking them based on how well they match what the user is really looking for. This ensures that the most helpful and accurate information appears at the top.

Broader Searches, Better Results

With reranking, we can search a larger set of data without losing accuracy. We retrieve a wide range of possible matches with vector search, then use the reranker to sort and return only the top results—usually the three most relevant. This focused approach improves accuracy and prevents unnecessary or distracting information from being included in our AI agents’ context window.

Benefits for All Our AI Solutions

By using reranking across our platform, all our AI agents, services, and processes get better, more precise information. Whether it’s an automated process, an assistant, or a data-driven service, reranking helps deliver clearer and more useful results, supporting effective decision-making and avoiding information overload.

Continuous Improvement for Our Customers

By adding reranking, we improve the quality and accuracy of the information our AI agents and services deliver. This helps our customers get clearer, more useful answers that support faster and better decisions.

We are committed to regularly reviewing and refining our technology so our solutions continue to meet our customers’ needs and keep up with the latest advancements in AI. By always looking for ways to improve, we ensure our platform stays effective, reliable, and valuable for the people who rely on it every day.

While our product development is driven by customer needs — with a clear focus on delivering meaningful value and a great user experience — some useful refinements also come from our own internal workflows. By using the same tools as our customers, we occasionally uncover issues or areas for improvement that might otherwise go unnoticed. These often take the form of enhancements and technical tweaks that help make the product more consistent, efficient, and pleasant to use. Here are just a few examples from that deeply technical, developer-shaped layer of our product work:

Service Integration Testing

We run regular integration tests for our services using Workflows in the Automation App and Flows in the Flow App. These tests help ensure that critical systems are online and communicating — but more importantly, the process of building them often exposes issues that wouldn’t surface through automated tests alone. One early example involved timer activities and the background jobs behind them, which helped shape our current approach to recurring, time-sensitive workflows.

As we continued building and scaling these flows — often involving dozens of steps, conditions, and service interactions — we began encountering another category of challenges: friction in the UI and overall user experience. These are the kinds of issues that only become apparent when you’re deep in real usage. During one of these builds, a developer noticed that a field could really benefit from a format hint, quickly added it, and later received appreciative feedback from others who no longer had to guess the expected input. It was a small tweak, born out of hands-on use, that quietly improved the experience for everyone.

Deployment Helpers

Deployments are smoother thanks to internal tools we built using Automation. These workflows scan application changes in our repositories and ensure they move correctly through staging and production. While setting this up, we realized our Automation App needed to support a broader range of activities — particularly around interacting with the Flow App’s API. So we extended its capabilities.

This improvement didn’t just help us streamline our deployment processes — it made the Automation App more versatile for everyone. Customers gained access to new, more powerful building blocks for their own flows, and interestingly, even our own developers began using certain linked features more actively. Once those activities became easier to integrate, adoption naturally followed. What started as an internal need ended up unlocking better workflows across the board.

Weekly Reports & Development Diaries

We also use Flows for team reflections and weekly reports — a simple but valuable ritual. Filling these out regularly give us firsthand experience with the UI and revealed usability issues in the Flow App that wouldn’t show up in automated tests. In one instance, a developer stumbled upon a layout issue that only appeared with a specific type of real-world content — something our test cases hadn’t included. It turned out to be a fairly impactful UX flaw, and thanks to that internal discovery, we were able to fix it before it ever reached customers. These small improvements, driven by actual use, make the product feel noticeably smoother and more reliable for everyone.

These are just a few examples from the tip of the iceberg — many other cases are either too boring or a bit too revealing for a blog post . But the principle stands. Using what we build isn’t just a habit — it’s part of our product DNA. It keeps us honest, sharp, and empathetic. Because if we don’t feel good using it ourselves, we can’t expect others to either.

Jetson Orin Nano: Strong but Unstable

We tested the NVIDIA Jetson Orin Nano under real-world maritime conditions to assess its readiness for deployment on vessels. It showed good performance when handling straightforward tasks like analyzing camera feeds or monitoring environmental inputs. The device offered a compact, capable AI platform that, at first glance, seemed well-suited for space-constrained setups.

However, during extended testing, a number of serious limitations became clear. The setup process was complex, involving multiple steps that were easily broken by system updates. Keeping the system running reliably required manual fixes, as the documentation was often missing or outdated. The unit also proved to be unstable under moderate load, with unexpected reboots even when active cooling was in place.

Attempting to run more complex AI models (for example, those used in decision-making or basic AI assistance) quickly pushed the device to its limits. Memory became a bottleneck, and compatibility issues emerged when installing newer software components. These challenges underscored the fact that the device lives up to its classification as a developer kit—suitable for testing and experimentation, but not ready for dependable deployment in real-world maritime operations.

In summary, while the Orin Nano is a powerful tool for very specific and lightweight tasks, it lacks the maturity required for unattended, production-grade deployment at sea. We consider it a promising platform for prototyping rather than one ready for operational use.

Verdict: Good for experiments and demos. Not ready for real operations at sea.

Google Coral TPU: Simple and Reliable

The Coral TPU by Google is designed specifically for running small and efficient AI models at the edge. It works well for detecting patterns in images or sound, making it a great option for fixed-function tasks like identifying equipment states, spotting safety risks, or monitoring for predefined anomalies.

One of its biggest strengths is how quickly and reliably it can be deployed. The setup process is straightforward and doesn’t require deep technical knowledge, which makes it appealing for teams looking to integrate AI without committing extensive engineering resources. Once deployed, it runs smoothly with very little need for maintenance.

That said, the Coral TPU is limited in what it can do. It only supports specific types of models, and those models must be pre-trained and optimized to run on Google’s TensorFlow Lite framework. It’s not built for flexibility or for running more interactive or complex AI logic, such as checklists or support tools that need to respond to changing input.

In short, it’s a stable and efficient solution for predefined detection tasks but not suitable for anything that requires broader AI reasoning or adaptability.

CPUs: Still the Best for Smart Agents

When it comes to running AI systems that help with decisions, provide instructions, or offer onboard support without an internet connection, regular computer processors (CPUs) remain the most reliable and accessible solution.

Unlike specialized chips designed for specific tasks, CPUs are general-purpose and can run a wide variety of software without special setup. This makes them ideal for smart agents that guide crew through checklists, assist with troubleshooting, or adapt to different situations. These agents do not require heavy graphics processing or large amounts of memory—just enough computing power to interpret instructions, respond to events, and maintain consistent behavior.

Another major advantage is that CPUs operate in a stable and familiar environment. Most operating systems and deployment tools are already built for CPU-based systems, eliminating the need for low-level configuration or hardware-specific patches. This significantly reduces setup and maintenance effort, which is especially valuable on vessels where technical support is limited.

In our own tests, CPUs were the only option that could reliably run basic AI agents without needing workarounds or encountering resource constraints. With the use of smaller, optimized models, it is entirely possible to deploy helpful onboard tools that operate independently and support the crew in real time, even in completely disconnected environments.

A further benefit of CPU-based systems is their ability to share resources across multiple services. By running everything in a virtualized environment, we can dynamically allocate CPU power where it is most needed, helping to smooth out performance during temporary demand spikes and making efficient use of available processing capacity.

BitNet: New Tech Worth Watching

BitNet is an experimental AI model architecture developed by Microsoft that focuses on extreme efficiency. Unlike traditional AI models, BitNet uses a technique called 1-bit quantization, which dramatically reduces how much memory and processing power the model needs to operate. This makes it especially promising for small, general-purpose processors like those found in many embedded and onboard systems.

While still in early development, BitNet represents a potentially important breakthrough for maritime use cases. With this kind of model, it might become feasible to run basic AI agents such as voice-assisted checklists or maintenance helpers directly on compact devices, without requiring a cloud connection or specialized hardware. It is not intended for large-scale decision-making or deep reasoning, but for simple and interactive onboard tools, it could prove to be a game-changer.

Currently, BitNet is still evolving. Only a few test models have been released, and the ecosystem is just starting to take shape. There’s still work needed before it’s ready for production use. However, the direction is exciting and aligns well with the needs of isolated environments like vessels.

Verdict: BitNet is a technology to watch. It’s not ready yet, but it holds strong potential for enabling lightweight AI assistants in future offline maritime systems.

Summary

There is no single solution that fits all scenarios when deploying AI at the edge, especially in maritime environments. Each platform we evaluated (Jetson, Coral, CPU-based setups, and even early-stage models like BitNet) offers different strengths and trade-offs depending on the use case.

Some devices are well suited for handling visual monitoring or predefined detection tasks, while others shine when used for logic-based agents or interactive tools. The key takeaway is that choosing the right approach depends heavily on the specific goals and technical context of the deployment.

The space is evolving quickly. New lightweight models, improved hardware, and better software support are emerging all the time. For that reason, we recommend reevaluating available options when planning any future AI deployment onboard. What isn’t feasible today might become viable in the near future.

In this post, we’ll share what the assistant has learned since its debut, what new AI agents are joining the crew, and give you a quick peek at a new tool we’re building to help you understand where your data actually delivers value.

But let’s start with our star player.

The AI Assistant Grows Up (and Gets a Memory)

The assistant has been hitting the gym. It now has long-term memory—but not in a confusing, “remember everything forever” kind of way. Instead, it remembers what you’ve discussed for the workflow you’re currently working on. So if you’re building a checklist workflow today and a sync process tomorrow, each one has its own focused memory. No mixed-up ideas, no overlap, just clean, contextual conversations.

Even better, the assistant now understands the most commonly used activities in the platform. That means it can help build more powerful automations and even comment on workflows you’ve already created. It’s learning to give advice, not just follow instructions.

We’re also wiring it into the app itself. Once that’s done, it’ll be able to proactively suggest improvements while you’re building—like a helpful co-worker who doesn’t need coffee breaks.

And yes, it now knows who you are. Thanks to OpenID integration, the assistant can greet you by name, tell users apart, and make sure only the right people have access. It’s polite and secure.

Right now, the assistant is undergoing internal user acceptance tests. Our own team is putting it through real-life scenarios, pointing out where it’s helpful and where it still needs polish. Once we’re happy with the results, we’ll roll it out gradually—starting with early adopters under a feature flag.

New AI Agents: Smarter Onboarding, Less Manual Work

While the automation assistant is learning to be your co-pilot, we’ve been training another AI agent to help you get started faster—especially when you’re facing a pile of paper checklists.

Let’s be honest: converting manual checklists into digital workflows isn’t fun. It’s repetitive, it’s slow, and it’s easy to miss small details that later cause big problems. That’s exactly where our new checklist digitization agent steps in.

This agent works alongside you to transform your existing checklists into clean, structured templates—fast. But it’s not just a digital scribe. It understands context and can recommend better controls along the way. If you’ve written “engine oil level (OK/LOW/HIGH)”, it won’t just drop in a text field—it might suggest a dropdown, or even a lookup connected to your system.

And it gets better. The agent automatically handles things like timestamps, user actions, and dates so you don’t have to create fields for those. It helps speed up input by proposing options or filling values based on past data. If something looks off—say a temperature way outside normal ranges—it can raise a flag. That’s not just convenience; it’s safety and quality.

This makes your checklists:

Faster to fill out

More accurate

Easier to maintain

Smarter out of the box

The best part? You stay in control. The agent supports you, but doesn’t replace your judgment. It’s there to reduce the busywork, so you can focus on the tasks that matter.

Coming Soon: Know What Your Data Is Really Doing

Behind the scenes, we’re working on something new—a tool designed to answer a simple but powerful question:

“What’s the actual value of the data I’m collecting?”

We’re calling it the data point tracking system, and it’s being built to give you a clear view into how your data flows across workflows, reports, and integrations. You’ll be able to see where each data point is reused, how often it’s accessed, and where it contributes to something meaningful—like a decision, a report, or an automation step.

Think of it like a radar for operational insight. You’ll spot the metrics that matter most, the ones that are just noise, and find ways to improve your processes based on actual usage—not just guesswork.

We’re still in early development, so we’ll save the deep dive for another post. But we wanted you to know it’s coming—and it’s being designed with the same focus on clarity, usability, and real-world impact.

Wrapping Up

From long-term memory in your automation assistant to checklist digitization and upcoming insights into your data, everything we’re building right now has one goal: to make your life easier.

Less repetition. Fewer mistakes. Faster onboarding. Smarter tools that understand what you’re trying to do—and help you do it better.

As always, we’re building this with you, not just for you. If you’re part of our early adopter group, we’d love to hear your feedback. And if not—stay tuned. There’s more coming soon.

“Just Make Me a User Creation Step”

Let’s say you type:

“I want to create a new user.”

Boom — you get a pre-filled HTTP activity ready to drop straight into your workflow. It knows how to talk to Maranics systems because it uses our Swagger definitions under the hood. No fiddling, no setup — just go.

 And yes, the generated step actually works — no smoke and mirrors.

Where This Is Going

Right now, we’re focused on generating individual activities. But the vision goes way further. Imagine saying:

“Set up onboarding for a new crew member.”
…and having the assistant build the entire flow — branching logic, conditions, integrations with external systems — everything.

Not Replacing You. Just Skipping the Boring Stuff.

We’re not trying to automate humans out of the loop. You’ll still want to tweak and customize the steps to match your real-world needs. But our goal is simple: let the assistant handle 90% of the boilerplate so you can focus on the work that actually matters.

Watch It in Action

We put together a short demo video of our first prototype. The user just types “create a user” — and the assistant instantly responds with a complete HTTP activity.

This isn’t a fake UI — but it is an early proof of concept. The activity is fully generated by the assistant, but it’s not yet available in the live product.