Before I Trust an AI Feature
What I want a product team to specify before an AI feature ships: expected behavior, failure modes, evaluation, and control.
Writing
Questions I keep coming back to.
Much of my recent writing is about AI: what a working demo leaves unanswered, how much authority a system should have, and what makes it useful in someone’s day-to-day work.
The Agentic AI Product Gap
What changes when an AI system can take action, and ten questions for working through autonomy, risk, evaluation, and human control.
From Hype to Value: Why Most AI Pilots Fail and How to Ship Real ROI
An essay on the work between an AI pilot and a useful product: data, workflow integration, evaluation, and a reason to keep investing.