Blog
Notes on building, evaluating and running enterprise AI agents.
- Why AI agents need evals, and how to build ones you can trust Evals show whether an agent does its task every time, and after every change. What the research says, how to build a suite you can trust, and what the evidence does not show.
- Why legal expertise matters when building enterprise AI agents An agent acts for the company that runs it, so the rules it acts under are requirements for the system, not a review before launch. Where those rules come from, how each one becomes a control, and what legal expertise does not replace.
