What the Board Actually Wants to See on AI

A board asks for an AI update and receives a demo, a model architecture diagram, and a slide of logos. None of it answers the question they were actually asking, which is some version of: are we exposed, and are we getting anything for the money. Boards do not need to understand your architecture. They need three things they can govern against: exposure, progress, and dependency. **Exposure** — what can our AI systems do without a human in the loop, and what is the worst outcome if one is wrong. **Progress** — what has moved from pilot to production since we last met, and what stalled. **Dependency** — which vendors and models are we now unable to switch away from cheaply. Each is answerable from data you already hold. None requires new tooling. All three are uncomfortable, which is why they are usually replaced with a demo.