Outcomes before capabilities
Useful AI starts with a consequential problem, an accountable owner, and a way to measure what changes.
AI product leader, builder, and writer
I lead AI product work inside a large, regulated enterprise — and build independent products to see where the technology is going before it gets there.
Read the field notes
01 / Current thinking
Useful AI starts with a consequential problem, an accountable owner, and a way to measure what changes.
The future is governed citizen development: more people solving problems within clear technical and human boundaries.
Read the case studyThe work changes when your judgment has to travel through people, portfolios, operating models, and culture.
02 / Writing
Essays about building useful AI, leading through uncertainty, and turning new capabilities into real systems.
Agent skills
Agentic engineering
Human-computer interaction
AI product quality
A personal detour
03 / Selected work
A visual home for saving, versioning, organizing, and moving agent skills between AI tools. Built from a personal need during the first wave of persistent, tool-using agents.
Two persistent agents run at home for months to find out where they earn their keep — and where agents end and deterministic software begins.
04 / About
I have spent 16 years inside a large, regulated enterprise, leading AI product teams across finance, supply chain, customer service, sales, and legal — and teaching rooms of 20 to 100 people what AI can and cannot do.
Outside the work, I am a husband and father of three. I make room for writing, hiking, biking, golf, Formula 1, and the occasional idea that refuses to leave me alone.
Currently on the shelf
Mik Kersten
Michael Connelly