Jared
Clark

The companies I work with answer to an auditor or an FDA investigator. I usually start with a gap assessment and stay until the certificate or the inspection report arrives. After twelve years I think most of that work comes down to one habit, which is being able to show who checked the work and where the record is.

Then AI showed up in those same companies. It began drafting procedures and filling in records, in some cases before anyone had decided who was supposed to check it. What happens when the auditor asks the old question about the new tool? When the model gets something wrong, who checked it and where is the record?

So my work now has two sides. One side is governance, which means ISO/IEC 42001 AI management systems and the AI governance programs around them. The other side is building. I build production AI systems for regulated companies, and I build them to survive an FDA inspection or an ISO audit. So far that has meant document control and training systems, and an ERP.

I think each side keeps the other honest. It is hard to write a sensible rule for a tool you have never built, and it is just as hard to build a tool for a plant whose rules you have never had to defend.

These are the degrees and certifications I hold. Where the issuer publishes a verification record, the name links to it.

I also write and build around a question that sits outside the consulting. How do communities hold together through shared truth-seeking rather than shared fixed beliefs, and how do people keep their agency as AI arrives? I have been asking some version of it for almost twenty years, and I don't expect to finish. That work has its own page.

Read the mission →