Ketryx AI Agents: How People and Agents Build Regulated Products Together

In regulated industries like life sciences and medical devices, quality isn't optional—it's safety-critical. But ensuring quality has often meant sacrificing speed and innovation. Product, engineering, and quality teams are stuck in time-consuming loops: assembling documentation, double-checking traceability, and manually verifying that every requirement, test, and risk has been properly addressed.

And it’s only getting harder. 

The rise of technologies like AI, robotics, and cloud computing, combined with evolving regulations and surging page counts in technical documentation and submissions, is amplifying the challenges faced by regulated teams building products of enormous complexity.

This is especially true as documentation burdens explode. The average medical device filing has ballooned from a few dozen pages in the 1980s to over 4,000 today—and could surpass 6,000 by 2028. Many of the complex products we support at Ketryx require 10,000+ pages of evidence to back up their claims. Manual compliance simply can’t keep up.

In an attempt to meet these challenges, organizations do what they’ve always done: scale headcount. But even with more people, it's difficult to build more complex systems and move faster when the underlying process is still disconnected, complex, and manual.

At Ketryx, we believe regulated teams shouldn't have to choose between speed and safety. That’s why we built Ketryx Agents—AI-powered coworkers designed to automate the most tedious and error-prone parts of regulated development, while increasing quality and safety.

Why Now? The Need for Accountable Autonomy

For the past decade, I’ve been working on implementing large-scale AI systems in enterprises, first with more traditional rule-based systems (like Wolfram|Alpha) and then on various AI systems, from NLP to image classification and generative AI. 

LLM-based chatbots like ChatGPT have opened the floodgates for AI, but they’re not ready to shoulder the responsibility of regulated enterprise workflows. What’s needed now is not just assistance, but accountable autonomy—systems that not only provide answers, but also take action within a controlled, validated framework that humans can trust.

Unlike chatbots, agents can perform routine tasks and take initiative: detecting risks, finding redundancies, performing root cause analysis, and generating documentation. But in safety-critical domains—where the margin for error is zero—agents must operate under strict oversight. Without this robust validation and human review, AI risks introducing more harm than help.

Validated AI agents are the answer: purpose-built AI teammates that operate within a validated, human-in-the-loop environment to scale your team, not replace it.

"Ketryx’s approach addresses the critical bottleneck of documentation and compliance that has historically hindered innovation in regulated industries," said Holger Mueller, Vice President and Principal Analyst at Constellation Research. "By automating the tedious aspects of documentation while maintaining human oversight and validation, Ketryx’s agents enable organizations to achieve both regulatory compliance and agile development - a combination that has been elusive until now. Companies that leverage these types of AI agents as an added layer of quality review will gain significant competitive advantages in time-to-market without sacrificing safety or compliance."

Moving from Lifecycle Management to Lifecycle Intelligence: The Rise of Zero-Lag Safety 

Ketryx was founded with a vision for zero-lag safety, where safety-critical issues are identified, investigated, and addressed as they happen, not months or years later. This eliminates the dangerous time gaps between when a problem occurs, when it’s identified, and when a solution is validated and deployed. Today, during this time gap, patients' lives are at risk. 

I know from firsthand experience that building, deploying and upgrading safety-critical systems is incredibly difficult, involving complex bureaucratic processes and thousands of pages of compliance paperwork. Many regulated teams employ a manual approach to compliance, painstakingly ensuring that every requirement, test, specification, and risk is double-checked and documented. But this process doesn’t scale at the speed of AI or with the increasing complexity of products today. This was the challenge we founded Ketryx to solve. 

To make this vision of zero-lag safety a reality, companies must be able to process bugs, emerging risks, and complaints continuously. They need to identify root causes in real time and make product changes without facing friction from manual documentation and compliance work. 

Achieving zero-lag safety requires embedding validated intelligence directly into the tools where work happens, empowering teams to document, trace, and approve work at the level that makes sense for the task at hand within their preferred systems. Only by transforming quality into an active, always-on, intelligent process—rather than a retrospective paperwork and compliance exercise—can companies accelerate innovation while maintaining the highest standards of patient safety.

We designed Ketryx as an AI platform to manifest this vision. Deterministic, rule-based, validated automation is used to ensure workflows are executed correctly, and artifacts are tracked and traced. Now, with the maturity of generative and agentic AI, we’re exposing the next leap forward. We have been using these features internally now for a number of years, and we’re excited to share them with other teams building regulated products.

‍Introducing the First Set of Safety-Critical AI Agents for Regulatory Compliance

The future belongs to teams who build with AI. In chess, humans collaborating with computers defeated both the best computers and the best humans — and in product development, teams that build products alongside AI will outperform those who don’t.

Meet the first set of Ketryx AI Agents: intelligent, semi-autonomous tools that execute critical tasks across your systems of work—Jira, GitHub, ADO, BitBucket, TestRail, and beyond. Every action they take is traceable, explainable, auditable and governed by your quality system. These aren’t generic chatbots—they’re purpose-built for safety-critical products, ensuring quality and helping you generate compliance artifacts from the work you’re already doing. In addition to these ready-to-use agents, teams can also create their own custom agents to run at a designated cadence. 

By leveraging Ketryx Agents, companies in safety-critical industries can reduce documentation time by 90% and achieve acceleration that is 10 to 100 times faster than the industry baseline. More importantly, products created with the help of AI agents are safer, as these agents can identify and perform critical tasks that a human might easily overlook.

Meet Our Six Ready-to-Use AI Agents

Requirement Conflict Agent

Review conflicting requirements 95% faster. This agent automatically detects and flags conflicting requirements so you can catch issues early, avoiding costly rework and reducing compliance risk. Learn more.

Redundancy Detection Agent

Identify redundancies 85% faster. Scans requirements, detects duplicate or overlapping items, and recommends fixes so that teams aren't uncovering duplication in final documentation or at the end of the release cycle. Learn more.

Test Coverage Agent

Scans tests and requirements to evaluate test coverage and surface gaps, so you can confirm that everything that needs a test has one and analyze test coverage 90% faster. Learn more.

Anomaly Review Agent

Review anomalies 90% faster. The agent verifies the completeness of anomaly records in your system (like checking for root causes), flags missing evidence, and suggests updates so that critical anomalies don't get overlooked. Learn more.

Change Request Review Agent

Verifies that every change request has all the appropriate fields filled out, flags any gaps, and suggests updates for completeness and consistency so you can reduce rework by 80%. Learn more.

Complaint Agent

Streamlines complaint intake, documents and categorizes complaints instantly, and flags critical signals so you can uncover quality issues early and triage complaints 80% faster. Learn more.

Each agent works inside the validated Ketryx platform. You can schedule them or trigger them automatically based on lifecycle events. All the work done by agents is shown with redlines and then reviewed by a human—nothing changes unless you approve it.

Watch this video to see the templated agents in action:

‍How It Works: AI-Driven Compliance You Can Trust

Despite their massive potential, generic, unvalidated AI agents also carry serious risks. Agents that are purely generative in nature rely on statistical models, which can lead to dangerous inaccuracies or hallucinations—a scenario that is unacceptable for safety-critical systems with patient lives at risk. 

Regulation and control frameworks already exist to help companies mitigate these serious risks, including IEC 62304 and ISO 14971 (co-authored by Ketryx’s VP of Regulation Paul Jones) and more recent reports focused on AI, like TIR 34971, which I was lucky enough to help write. Safety-critical industries can adapt and build upon these established principles to ensure the responsible deployment of AI agents. Based on our work building widely deployed, high-risk products with three of the top five medical device manufacturers, we've developed a risk-based framework for AI agent utilization. This framework is intended to help companies utilize AI agents in a way that accelerates time to market while keeping patients safe and humans in the loop. 

Ketryx Agents follow a repeatable, validated, reviewable process:

1. Prompt: Describe a task using everyday language.
2. Analysis: Ketryx Agents execute the task while analyzing your project across all connected systems to flag issues and gaps. Each analysis is in-depth and different in structure based on the context of the query.
3. Recommendations: Agents provide suggestions with redlines and clear explanations, making it easy for anyone to understand what is being recommended (and why) for each change. Agents also suggest which items should be traced and which items and documents need to be created or modified, enabling faster, more complete traceability. 
4. Human Review: Agents will never take any action; a human must review and approve every recommendation agents make. This means that you keep your experts in control of your process while making them more efficient. ‍
5. Sync: With your approval, agents will take action throughout your connected systems (like Jira), updating requirements and adding traceability where needed.

The Future of Validated AI for Regulated Teams

Companies simply cannot hire enough people to do the work they need. Validated AI agents give companies the opportunity to multiply their staff and move significantly faster. But this isn’t just about saving time or making compliance simple; it’s about improving the quality and safety of systems proactively, because AI never gets tired, never misses a step, always completes tedious documentation work on time, and always follows the rules.

These gains are not merely hypothetical. Companies utilizing computers to do tedious work like documentation and traceability are already witnessing gains in release speed that they never thought possible. With AI agents, that impact could go much further. Quality and compliance will become continuous and embedded in the product lifecycle so that we can reach a state of zero-lag safety. 

This isn’t just the right thing to do for patients; it’s also the right economic decision as companies could save 30% or more on project costs, allowing them to drive more innovation through their development pipeline faster. We’re excited to lead the way to this promising future where regulated teams collaborate with AI agents to release safety-critical software faster without compromising compliance.

See Ketryx Agents in Action

Want to see how agents detect and fill traceability gaps, check for conflicts in your requirements, or surface missing test coverage? Join our upcoming webinar to watch Ketryx Agents at work—and learn how teams are releasing faster, with more confidence and less risk.