Governed Autonomy

This book is for the person who decides. It gives you a model for governing AI agents, the questions to ask of the people who run them, and the evidence to defend the answers upward. Every chapter ends on a decision, because a decision is what an executive actually produces.

It is not an implementation manual, and it is deliberately not one. It has no code, no configuration, no architecture you can hand to a team, and no template to fill in. Four things are missing on purpose: templates, the build-versus-buy decision, detection engineering, and the org chart. If you need those, this is the wrong book. My previous one, Agentic by Design, is the practitioner's blueprint: the tooling, the pipelines, the ninety-day build.

This book will change what you ask for and what you sign. It will tell you which of a vendor's claims are doing work and which are decoration, what the law now requires of you rather than of your model provider, and how to tell whether the governance you have paid for is operating or merely documented.

Read it as a briefing you can interrogate, not a plan you can delegate.