Product Teams Are Disappearing. Here's What's Replacing Them.

The traditional software product team is dissolving. Where companies once needed specialized product managers, engineers, designers, and QA staff working in sequence, AI is collapsing that structure into something faster and flatter.

The shift started with code generation-leading tech companies report that 30 to 50 percent of their code is now machine-generated. But that's the visible part. The real change happens earlier and later in the development cycle.

The Death of the PRD

Long product requirements documents are disappearing in advanced organizations. Instead of writing exhaustive specifications, product managers now move directly to prototypes.

AI enables rapid mockups, fast iteration, and real-time testing without waiting for a full design or engineering cycle. The result is faster learning and quicker delivery to market.

This creates a new role: the product builder. Rather than coordinating between specialists, these product builders orchestrate AI agents across design, engineering, testing, and deployment. Teams deliver exponentially more output with fewer handoffs.

Commerce Without the Human Shopper

The implications extend beyond how teams work. They reshape what products need to do.

Retailers and marketplaces traditionally optimized digital experiences for human behavior-apps, websites, conversion funnels designed around how people click and buy. That assumption no longer holds.

AI agents now research, compare, and purchase on behalf of users. Consumers can initiate transactions through large language models. Agents negotiate and execute purchases independently.

This means retailers must optimize for two audiences at once: human shoppers and the AI agents that shop for them. Search becomes horizontal, with users' chosen LLM functioning as their shopping portal. Marketplaces face disintermediation risk. New winners will emerge based on how seamlessly they integrate into agent-driven commerce.

For some companies, this shift opens growth opportunities. For others, it's existential.

What Product Leaders Need to Know

Product development is no longer about building interfaces for humans alone. It's about designing systems that both humans and machines can navigate effectively.

Teams need different skills. The specialized roles that defined product development-front-end engineer, back-end engineer, designer-are being reconfigured. Product builders need to understand how AI agents interact with your product, not just how users do.

This affects hiring, team structure, and operating models. Organizations asking how fast they can ship differentiated digital experiences now are asking the right question. So are those asking who they actually need to hire in an AI-native world.

For product professionals, this means the toolkit is changing. AI for Product Development isn't a side skill anymore-it's central to the job. Understanding how to work with AI agents, prototype at speed, and design for machines alongside humans is becoming table stakes.

Amazon Projects AWS Revenue Could Double to $600 Billion by 2036

Amazon CEO Andy Jassy told employees this week that artificial intelligence could push AWS to $600 billion in annual sales within a decade - double his previous estimate of $300 billion. The projection emerged during an internal all-hands meeting Tuesday and suggests AWS would need to grow nearly 17% annually from its 2025 revenue of $128.7 billion.

Jassy framed the higher forecast around concrete demand. "We have very clear and significant demand signals," he said. "We're not just spending the $200 billion of capex because we're hoping AI is going to be big."

The company's planned $200 billion capital expenditure for 2026 - mostly for AI infrastructure - rattled Wall Street. When asked about the spending during the meeting, Jassy acknowledged the scrutiny. He explained that AWS must commit capital years in advance for data centers, power systems, chips, and networking equipment before revenue materializes.

What This Means for Sales Teams

For sales professionals, these numbers signal where enterprise spending is heading. A $600 billion AWS business would require massive expansion in cloud services, AI tools, and infrastructure solutions. Companies building sales strategies around cloud adoption and AI capabilities are positioning themselves for the growth cycle Jassy describes.

The revenue projection also reflects how AI is reshaping customer demand. Jassy said AWS sees "very unusual opportunity to build this very large business" - language that translates to expanded customer needs and longer sales cycles as organizations deploy AI infrastructure.

Learn more about AI for Sales or explore the AI Learning Path for VP of Sales to understand how these market shifts affect revenue strategy.

Other Business Updates

Amazon also announced it expects to reach one million drone deliveries this year. The program, in development since 2013, promises 30-minute delivery for items fitting in a shoebox.

The company closed its Fresh and Go store formats in January. Those physical locations accounted for less than 1% of Amazon's overall grocery sales, the company said during the meeting.

AI Healthcare Market to Reach $110.61 Billion by 2030

The artificial intelligence healthcare market will grow to $110.61 billion by 2030 from $21.66 billion in 2025, expanding at an annual rate of 38.6 percent, according to market research from MarketsandMarkets.

Healthcare providers are driving adoption. Hospitals and health systems account for the largest share of AI spending today, driven by budget increases aimed at improving care quality while cutting costs.

Where AI Is Being Used

Deep learning-a subset of machine learning-commands the largest share of AI tools in healthcare. Beyond that, hospitals are deploying AI across diagnosis and early detection, treatment planning, patient monitoring, pharmacy management, and administrative functions.

Asia Pacific is seeing the fastest growth. Demographic shifts, technology improvements, and rising investment in innovation are accelerating AI adoption across the region.

Who's Building These Systems

Major players include Microsoft, Siemens Healthineers, NVIDIA, Epic Systems, GE Healthcare, Medtronic, Oracle, Veradigm, and Google. These companies are pursuing mergers, acquisitions, partnerships, and new product launches to strengthen their positions.

What's Holding Adoption Back

Clinician resistance remains a barrier. Many medical practitioners hesitate to adopt AI-based tools. A shortage of skilled professionals who can implement and manage AI systems compounds the problem.

Lack of standardized frameworks for AI and machine learning also constrains growth. Without consistent standards, healthcare organizations struggle to evaluate and integrate solutions.

What's Driving Growth

The need for earlier disease detection is pushing adoption. The volume and complexity of health data are growing exponentially as digital health systems expand. Rising chronic disease prevalence is putting cost pressure on providers, making AI efficiency gains more attractive.

Helen Weisinger launches AI production consultancy to reshape creative workflows

Helen Weisinger, former chief marketing officer EMEA at Tag, has co-founded Creative Engineers, an AI production consultancy that helps brands handle creative work in-house or restructure their agency relationships.

The consultancy targets the gap between creative teams and production capabilities. Brands increasingly need ways to produce more content with existing resources, and Weisinger's firm positions itself as a bridge between internal teams and external partners.

Creative Engineers aims to address a practical problem: many brands lack the tools or knowledge to use AI effectively in production workflows. The consultancy advises on implementation rather than selling software.

For creatives, this shift matters. Production bottlenecks often constrain what designers and creative teams can attempt. Tools that streamline production-whether through AI or process redesign-directly affect what's possible within a deadline and budget.

Weisinger's background in marketing leadership at a major production company gives the venture credibility with both agency and brand-side clients. Her role overseeing EMEA operations means she understands the regional variations in how teams operate.

The timing reflects broader industry trends. Brands are testing AI in production pipelines, but many lack clear strategies for integration. Consultancies that help teams move from experimentation to actual workflow changes are filling a real need.

For those working in creative roles, understanding how AI for Creatives applies to production can open new possibilities in your work. Practical knowledge of these tools and processes is increasingly expected.

PropertyRadar 5.0 Adds Court Data and AI Tools for Real Estate Investors

PropertyRadar released version 5.0 of its lead generation platform, adding court-sourced data on divorces, probates, and evictions alongside new AI assistants and integrated marketing tools. The update positions the platform as the only one combining property search, multi-channel marketing, and distress scoring in a single interface.

The new Distress Scoring feature ranks properties by likelihood of distress, letting investors prioritize outreach. A new Distress Criteria filter lets users build lists that include any property showing one or more distress signals, which helps generate viable lists in smaller markets where individual data points are sparse.

PropertyRadar added Relatives Data to property records and created category-specific investor "Plays"-guided workflows that move users from identifying opportunities to launching marketing campaigns. The platform now integrates direct mail, email, phone, and online advertising tools alongside in-person outreach options.

The AI assistants help users build audience lists, create brand messaging, and develop marketing copy. Mark Hockridge, the company's CEO, said the updates address investor frustration with inaccurate data and wasted marketing spend.

"PropertyRadar is committed to helping our users show up to deals first," Hockridge said. "We recognized the frustration of inaccurate data and wasted marketing efforts, and decided to invest in giving investors what they actually need to close more deals: fresh and reliable data, powerful marketing tools, and AI assistants to help bring it all together."

The company is offering a five-day free trial for early access to version 5.0. PropertyRadar has served hundreds of thousands of businesses since 2007, focusing on small and local real estate firms.

For real estate professionals exploring how AI fits into their operations, resources like AI for Real Estate & Construction and the AI Learning Path for Real Estate Brokers provide structured guidance on integrating these tools into workflows.

Rain Launches AI Agent to Help Employees Manage Personal Finances

Rain, a financial health platform serving 3.5 million employees at over 1,200 companies, released an AI agent designed to help workers manage cash flow, cut expenses and build financial stability. The system works continuously, taking user-approved actions to improve financial outcomes rather than simply offering advice or educational content.

The platform combines real-time earnings data from employers with financial information employees connect themselves. That dual data source gives Rain context that standalone financial apps lack, according to the company.

How It Works for Employers

Rain operates in the background, monitoring each employee's financial situation to prevent cash shortfalls and build long-term security. The approach differs from traditional benefits programs that require workers to seek out help or learn on their own.

Employers including McDonald's, Marriott and T-Mobile use Rain. The platform has distributed over $4 billion in earned wages and tracks measurable outcomes in employee retention, productivity and workforce performance.

What HR Leaders Should Know

Financial stress directly affects workforce performance. Employees dealing with cash flow problems miss work, show lower productivity and leave jobs more frequently. An always-on financial agent addresses these issues without requiring workers to remember to use it or understand complex financial concepts.

Rain's CEO Alex Bradford said the system "levels the playing field" by giving employees access to financial expertise previously available only to wealthy individuals. "Within a few years, everyone will have an AI financial agent," he said. "Rain will be the one they trust."

Existing customers can contact their Rain representative. New employers can request a demo to see how the platform integrates with their benefits offering.

For HR professionals exploring AI applications in workforce management, understanding how AI agents can improve employee financial wellness connects to broader talent strategy. AI for Human Resources covers this expanding field, while an AI Learning Path for CHROs addresses how senior HR leaders can evaluate and implement AI-driven employee benefits programs.

Nvidia Plans Chinese Version of Groq AI Chips After $17 Billion Acquisition

Nvidia is developing a China-compatible version of Groq AI processors it acquired last year for $17 billion, according to two industry sources. The company announced the move this week at its developer conference in San Jose, California.

The chips will handle inference operations-the part of AI systems that responds to user queries and generates code. Nvidia plans to pair them with its Vera Rubin chips, though those cannot be exported to China due to U.S. restrictions.

Nvidia CEO Jensen Huang separately announced the company has resumed manufacturing its H200 processors, an older generation chip, after securing Trump administration export permits and receiving purchase commitments from Chinese buyers.

What This Means for Sales Teams

For sales professionals, this signals Nvidia's aggressive push into inference-a market where it faces real competition. Chinese companies like Baidu have already built their own inference technology, so Nvidia can't rely on dominance in this segment the way it does in AI training.

The May availability date suggests Nvidia expects near-term demand from Chinese customers. Sales teams should understand that generative AI and LLM infrastructure decisions increasingly depend on inference capabilities, not just training power.

This also reflects broader geopolitical realities. U.S. export controls create fragmented markets, which means different regions now need different chip strategies. That fragmentation affects how enterprises plan their AI for sales deployments and infrastructure investments.

Key Details

• The China-bound Groq chips are not reduced-capability versions-they're standard chips configured for alternative system integration
• Expected availability: May
• Nvidia did not comment when asked about the development

OpenAI to sell AI models to U.S. defense agencies through Amazon cloud

OpenAI has signed a deal to provide its AI models to U.S. defense and government agencies via Amazon Web Services for both classified and unclassified work. The arrangement supports a Pentagon contract OpenAI secured last month after the military dropped its previous supplier.

The shift marks a significant change in how AI vendors access federal contracts. OpenAI previously focused on unclassified government use. Now it will handle classified military and intelligence operations-a market segment that requires deep integration with existing federal cloud infrastructure.

Why AWS matters

Amazon's cloud unit already operates within Pentagon systems, making it the natural distribution channel for government AI. Access to these embedded relationships is becoming a key competitive advantage in defense contracting.

OpenAI's partnership with AWS also reflects a broader shift. After transitioning to a for-profit structure last fall, OpenAI updated its agreement with Microsoft to allow partnerships with rival cloud providers when selling to national security customers.

The Anthropic collapse

This deal comes after the Pentagon's relationship with Anthropic fell apart. Anthropic had won a defense contract worth up to $200 million in July 2025 and was working with Palantir and AWS to deploy its Claude models in classified systems.

In February, Anthropic refused to allow unrestricted military use of its AI, particularly for domestic surveillance and autonomous weapons. The Pentagon labeled the company a "supply chain risk" and effectively cut it off from government work.

Government contracts as market signal

High-stakes public sector work signals trust and reliability to corporate clients. Securing Pentagon contracts could help OpenAI attract large enterprises that view federal approval as validation of security and capability.

For government professionals, these changes mean AI for Government work will increasingly center on ChatGPT and similar models deployed through major cloud providers.

What Dogs' Word Buttons Reveal About AI and Animal Minds

Federico Rossano's research on dogs using soundboard buttons to communicate has drawn millions of viewers to a recent NOVA documentary. But the work extends far beyond viral videos. His findings suggest that button-trained dogs learn a shared communication system with humans, responding to recorded words as consistent cues and, in some cases, combining buttons in ways that appear purposeful rather than random.

Rossano, an associate professor of cognitive science at UC San Diego and director of the Comparative Cognition Lab, paired controlled experiments with one of the largest citizen-science datasets in animal communication history-gathered from thousands of dogs in real homes worldwide. The research has shifted how scientists approach studying animal minds.

Why Labs Miss What Homes Reveal

Traditional animal research isolates subjects in controlled lab settings, often creating stress that skews results. Rossano's approach moves beyond that model. His team travels to homes, tests dogs with their owners present, and compares performance across different contexts.

"Many scientists refrain from citizen science because the claim is that we cannot trust what people do at home," Rossano said. "But I have found that engaging with the community leads to exciting insights that would be completely lost if we refrain from engaging with the people who actually live with pets every day."

The shift reflects a broader change in how science operates. Modern tools can analyze messy, large-scale data in ways that weren't possible decades ago. Real-world observations now complement controlled experiments rather than replace them.

What Dogs' Buttons Tell Us About AI Understanding

The parallel between how dogs learn button patterns and how large language models work is direct. Both start by learning statistical associations: a dog learns that pressing "treat" produces a reward; an LLM learns which words typically follow others based on training data.

But there's a critical gap. "Current LLMs are learning statistical regularities," Rossano said. "The issue is that we do not quite know to what degree AI understands 'meaning' because they do not have 'world knowledge.'"

A two-year-old asking for a credit card after watching a parent buy toys demonstrates the problem. The child produces the correct word in context but lacks understanding of what a credit card actually is. AI systems face the same challenge at scale.

For researchers studying animal communication with AI tools, the stakes are high. AI excels at identifying patterns-which whale songs exist, how often they occur, what follows them. But without knowing what those signals actually mean, researchers can only predict sequences, not interpret content.

Mapping Animal Cognition Beyond IQ

Understanding what animals know matters for welfare and conservation. Knowing whether a species notices when a member goes missing, how long they remember, and whether they can communicate pain or emotional states changes how humans should treat them.

Some animals also detect environmental changes humans miss. Dogs' sense of smell is unmatched. Certain species show early signs of earthquakes or wildfires. If researchers can decode what animals perceive and communicate, humans could benefit from that knowledge.

"After all, we have built airplanes because humans could not stop being mesmerized by birds flying," Rossano said.

The Next Frontier: Primate Behavior Models and Wild Animal Tracking

Rossano's lab is developing "Primate-GPT," a foundational model using computer vision and multi-modal AI to track primate behavior and communication in context. Understanding who does what, with whom, and when reveals patterns that improve animal communication research.

For conservation, the work addresses a practical problem: monitoring wild animals without invasive methods like darting. Tracking relationships between individuals matters. A primate is more likely to follow family or close associates than strangers-just as a human would.

"If we understand who they are to each other, it is an important predictor of who they will follow and what they will do next," Rossano said. "Tracking relationships and behavior over time is a new frontier for AI for conservation."

Current Research Questions

Rossano's team is investigating whether dogs and cats can reliably communicate about pain, emotional states, and needs for help. They're also testing whether animals can combine buttons to express concepts without direct word equivalents-a skill called "productivity" that humans demonstrate constantly.

If dogs can grow their communicative signals to match changing environments, it would suggest they experience the world in ways more similar to human thinking than previously understood. That finding would also address open questions about animal sentience and consciousness.

For professionals working in AI for Science & Research or studying how generative AI and LLM systems process meaning, the implications are direct: animal communication research offers a testable framework for understanding what AI systems actually know versus what they merely predict.

Trustero AI Overhauls Compliance Evidence Management With Multi-Agent System

Trustero AI announced a redesigned Evidence Management system that automates compliance evidence collection, control mapping, and analysis. The company said the changes will save organizations hundreds of hours annually by replacing manual processes with AI-powered workflows.

The problem is straightforward: compliance evidence sits scattered across SharePoint folders, cloud environments, ticketing tools, and dozens of other systems. Compliance teams collect it manually, upload it once, and leave it static. Traditional GRC platforms offer little help.

Three Core Capabilities

Trustero's system adds three features designed to address this friction:

• Automated collection. Trustero's Receptors pull evidence from AWS, Jira, and other platforms. A new folder scan capability syncs entire directories from SharePoint and Google Drive, eliminating manual uploads.
• AI-powered mapping. The system analyzes evidence and recommends which controls it matches, removing the need for manual mapping and reducing audit risk from mismatched evidence.
• Interactive analysis. A new copilot tool lets compliance teams query evidence using natural language. It can combine multiple pieces of evidence, perform row-by-row analysis on tables, run random sampling, filter large datasets, and save workflows as reusable reports.

The system also versions every piece of collected evidence and automatically surfaces the version applicable to each audit's timeframe.

Measurable Impact

Trustero customers are already automating monitoring across 112 high-risk controls, according to the company. A centralized evidence repository also serves as a foundation for automating adjacent processes like security questionnaires.

Phil Liu, CEO of Trustero AI, said: "Compliance teams have been stuck managing evidence the hard way for too long. Scattered repositories, manual uploads, and static data are audit risks hiding in plain sight."

For managers overseeing compliance functions, the core value is straightforward: reduce manual work, lower audit risk, and extract actionable intelligence from evidence already being collected. Learn more about AI Agents & Automation for Compliance Workflows and AI for Executives & Strategy.

Surf AI Launches $57 Million Security Operations Platform

Surf AI, a New York-based startup, announced its launch Tuesday with $57 million in funding to help operations and security teams close exposure gaps at scale. Accel led the round, joined by Cyberstarts and Boldstart Ventures.

The platform connects fragmented security data across identity, cloud, HR, and IT systems to build a context graph showing assets, owners, permissions, and dependencies. It then prioritizes risks by business impact and coordinates remediation through automated workflows.

How It Works

Surf AI uses specialized AI agents to drive execution while keeping humans in control. The system preserves context as work progresses, eliminating repeated handoffs and manual rework that typically slow security operations.

Yair Grindlinger, CEO and co-founder, said the platform addresses a persistent operations challenge: "Proactive security hygiene is exactly what we're encouraging, and our platform is designed to continuously find and close the exposure gaps that teams have always known about but didn't have the time or resources to address."

Team and Traction

The company was founded in 2024 by Israeli cybersecurity veterans including Grindlinger, Elad Horn, Roie Cohen Duwek, Avner Gideoni, and Brenton Gumucio. Surf AI is already working with Fortune 500 companies and global organizations.

The funding will support product development and hiring. Operations leaders managing security remediation workflows may find AI learning resources for cybersecurity operations helpful as these tools become more prevalent in SOCs.

Lighthouse Guild Launches AI Initiative With Blind Leadership at the Center

Lighthouse Guild announced the launch of Lighthouse Guild AI (LGAI) on March 17, a new division dedicated to co-developing adaptive technology with partners like Meta. The organization places blind and low-vision people in leadership roles during technology design, not as testers after development.

The announcement came as Thomas Panek, the organization's CEO who is blind, completed the United Airlines NYC Half marathon on March 15. Panek ran 13.1 miles through New York City using AI glasses co-developed with Meta, with real-time coaching from ultramarathoner Scott Jurek through the device.

How the Technology Worked

The AI glasses provided Panek with information about upcoming water stops and the location of the finish line before he reached it. A human guide, Jed Laskowitz, ran alongside him. The setup demonstrated how the technology could function in a real-world scenario rather than a controlled lab environment.

"Every time I've stood at a start line, people have told me what I couldn't see," Panek said. "What I've learned is that the real limits aren't in our eyes. They're in our tools and in our imagination."

Initial Focus Areas

LGAI will concentrate on three capabilities: navigation, travel, and obstacle detection. The organization argues these areas have the potential to increase independence and mobility for millions of people with vision loss.

Alex Himel, VP of Wearables at Meta, said the partnership reflects a basic principle: "Technology works best when you build it with the people who'll actually use it."

What This Means for Development Teams

For IT and development professionals, the model presents a different approach to accessibility work. Rather than treating vision loss accommodations as a separate feature set added late in development, LGAI embeds affected users into the design process from the beginning.

This approach mirrors broader shifts in how companies build for specific user groups. Involving end users early typically surfaces problems that internal teams miss, reduces rework, and produces features that actually solve real problems rather than assumed ones.

Jim Dubin, Chairman of Lighthouse Guild, said the race was a milestone but represented something larger. "Lighthouse Guild AI marks the beginning of a new chapter - one where our organization is both serving the blind and low vision community and actively shaping the technology that will define their independence for generations to come."

Learn more about Lighthouse Guild and its work in adaptive technology.

Dedicated AI Platforms Are Displacing Search as Discovery Starting Point

More than six in ten U.S. consumers used dedicated AI platforms in 2025, with some reducing their reliance on traditional search engines, according to a December 2025 report by PYMNTS Intelligence. The shift marks a structural change in how people discover and purchase products online.

The report surveyed 2,100 U.S. adults and found that consumers are moving away from the linear sequence of search, browse, compare, and purchase. Instead, they're using repeated AI prompts to handle all these tasks within a single platform-often without visiting a vendor's website.

Who's Adopting AI First

Gen Z and power users (those performing at least 25 AI tasks monthly) are driving adoption. Only 14% of light users report comfort using AI for banking tasks, while a third of power users and Gen Z are confident enough to use AI for personal financial decisions.

Older adults tend to avoid dedicated AI platforms, citing concerns about personal data and accuracy issues. These barriers are particularly significant in higher-stakes domains like personal finance and online payments.

Dedicated Platforms Change Behavior; Embedded AI Extends It

The distinction between dedicated AI platforms and AI embedded in search engines matters. Users of dedicated platforms rely 27% less on search engines than those using AI summaries within search results.

Among dedicated platform users, 43% report replacing their older search engine-based discovery methods entirely. By contrast, 59% of people using AI summaries in search say those tools complement rather than replace traditional search.

For marketers, this means the online environment is the critical variable. Brands may need to duplicate product messaging on AI systems to remain visible to consumers who've shifted their discovery habits.

Digital Wallets as Trust Layer

Consumers show a strong preference for digital wallets when making purchases through AI platforms. This preference likely reflects a desire to maintain control over payment data and rely on established authentication methods.

The report suggests digital wallets may become the mainstream trust layer in AI-mediated commerce, offering a middle ground between convenience and consumer control.

What's Blocking Wider Adoption

Privacy concerns, AI's difficulty understanding user intent, and inaccurate responses are the main barriers to adoption. Light users are most likely to use AI only for writing and shopping discovery-the lowest-risk applications.

Mainstream users prioritize reliability and safety, making them less likely to integrate AI into purchase decisions. This suggests adoption will remain segmented by user type and use case for the near term.

Traditional and AI Systems Will Run in Parallel

The report concludes that some consumers are moving toward AI-first navigation, but trust and accuracy issues are limiting broader adoption. Early displacement of search and app-based discovery is notable only among specific cohorts and in particular use cases.

For marketing decision makers, the shift is active in some segments. Traditional and AI-based discovery systems will likely operate in parallel for the foreseeable future. Brands that adapt to AI for Marketing environments may gain an advantage, depending on product type and target audience.

The report does not predict how quickly any transition will occur, but indicates the change is already underway.

NVIDIA Releases Dynamo 1.0, Inference Software Adopted Across Cloud Providers and Enterprises

NVIDIA released Dynamo 1.0 on March 16, open source software designed to manage AI inference workloads across data centers. The software integrates with major cloud providers including Amazon Web Services, Microsoft Azure, Google Cloud and Oracle Cloud Infrastructure, along with adoption from enterprises like ByteDance, PayPal and Pinterest.

Dynamo functions as a distributed operating system for what NVIDIA calls "AI factories" - data centers running inference at scale. It coordinates GPU and memory resources across clusters to handle requests of varying sizes and priorities arriving unpredictably.

Performance Gains and Technical Approach

In industry benchmarks, Dynamo boosted inference performance on NVIDIA Blackwell GPUs by up to 7x, according to the company. The software reduces token costs by managing memory more efficiently and routing requests to GPUs that already contain relevant data from earlier processing steps.

The system adds traffic control mechanisms and moves data between GPUs and lower-cost storage to reduce wasted computation. For long-context requests common in agentic AI systems, Dynamo can offload temporary memory when it's no longer needed.

Ecosystem Integration

NVIDIA integrated Dynamo with popular open source frameworks including LangChain, SGLang and vLLM. The company also released standalone modules like KVBM for memory management and NIXL for GPU-to-GPU data movement.

Cloud infrastructure providers CoreWeave, Nebius and Together AI said the software reduces deployment complexity. Chen Goldberg, executive vice president at CoreWeave, said supporting Dynamo allows the company to offer "a more seamless, resilient environment for deploying complex AI agents."

Who's Using It

Adoption spans multiple categories. AI-native companies Cursor and Perplexity use the platform. Inference endpoint providers including Baseten, Deep Infra and Fireworks integrated it. Global enterprises including AstraZeneca, BlackRock, Coupang, Instacart, Meituan, Shopee and SoftBank Corp. deployed it.

Pinterest said in a statement that Dynamo optimization helps the company "expand the seamless and personalized experiences we deliver, powered by high-performance AI infrastructure" to hundreds of millions of users.

Dynamo 1.0 is available now as free, open source software. Operations teams managing AI infrastructure should consider how inference orchestration affects GPU utilization rates and token costs. Learn more about AI for Operations and explore the AI Learning Path for Operations Managers to understand how these systems integrate with broader operational strategy.

Salesforce embeds AI into SMB CRM to speed customer action

Salesforce is bringing AI directly into Salesforce Suites, its CRM platform for small businesses and startups. The goal: help lean teams turn customer data into action without switching between systems to find context.

Small business teams often wear multiple hats. Sales, marketing and service responsibilities frequently sit with the same people, and toggling between tools to understand customer situations slows everything down. The new AI capabilities aim to close that gap by surfacing key information instantly.

Instead of searching across tabs for account history, deal status or support case details, teams can ask the system for a summary and get the information immediately. That shift cuts the time between understanding a situation and acting on it.

Small businesses are adopting AI faster than larger organizations

Small businesses are already moving on AI adoption. Research by Constant Contact found that 54% of small business owners use AI marketing tools, with another 27% planning to start in 2026.

Smaller organizations are integrating AI more broadly across their tech stacks than their larger counterparts. The State of Martech 2025 report found that 34% of SMBs have adopted AI across multiple workflows or fully integrated it into their martech stack. By comparison, only 14% of mid-market companies and 27% of enterprise organizations report that level of adoption.

Lean teams need tools that simplify daily work rather than adding complexity. Smaller companies often move faster because they must.

AI inside the CRM reduces friction in daily work

Sales teams can now generate instant summaries of deals that highlight recent activity, potential risks and recommended next steps. AI can also draft follow-up emails using existing customer data and conversation history, giving teams personalized drafts in seconds instead of starting from a blank screen.

Salesforce is also introducing an Employee Agent-a conversational assistant that can summarize records, draft communications and automatically log activities. When AI works directly inside the system where customer data already lives, teams move from information to action faster.

Workforce orchestration platform Asymbl reports that its sales teams previously spent about 15 minutes preparing for client calls by navigating multiple records and notes. Now they can request account context and immediately see deal status and recent activity.

The marketer's role is shifting toward orchestration

AI adoption introduces new challenges for marketing teams. Research suggests that 39% of mid-market marketers feel they lack the skills needed to adopt AI effectively.

The emerging role for marketers is shifting from manual execution toward orchestration. Teams increasingly need to set guardrails, interpret AI outputs and ensure automated decisions align with business goals. Marketers are becoming systems orchestrators rather than simply campaign operators.

Embedding AI directly into platforms teams already use may help reduce some of that complexity. When intelligence is built into familiar tools, the learning curve is shorter and adoption is easier.

For more on AI for Marketing and AI for Sales, explore resources tailored to your role.