This is my AI governance writing, one question per piece, with the sources it rests on. If a claim needs a number I cannot show you, it is not here.
· 14 min read · AI operating model
In METR's 2025 trial, 16 developers believed AI made them 20% faster. Tasks took 19% longer with AI allowed. METR now marks that result out of date.
production-readiness · ai-governance · shadow-ai · ai-pilots
· 12 min read · AI for boards
How to hand over AI leadership in steps that end in a test: what moves, who takes the role and how to tell it worked.
· 15 min read · AI operating model
How to choose an operating model for AI work in a mid-size company, starting from what the law requires and what the evidence supports.
· 15 min read · Agentic architecture
What the sources say about multi-agent systems in production: where vendors agree, where they split, and what the research did not find.
· 6 min read · AI governance
What a working AI acceptable use policy has to decide, what it must refuse to do, and how a company of 30 to 300 people gets staff to follow it.
acceptable-use-policy · ai-governance · shadow-ai · policy-enforcement
· 8 min read · Agentic architecture
How to build an eval set from production traffic, score an LLM agent without a golden answer, and fail a merge in CI when it regresses.
· 5 min read · AI governance
A step-by-step way to build an AI system register in a week without buying a tool, so a security reviewer gets the document they ask for first.
· 6 min read · AI operating model
Most AI pilots stall on ownership, data rules and a missing outcome metric, not the model. Here is what has to be true before one reaches production.
ai-pilots · production-readiness · ai-governance · evals-in-ci