Give the board a shared language for AI decisions.

Understand capability and limits

Distinguish systems that classify, predict, generate, recommend and act. A convincing output is not evidence of correctness, and a model's explanation is not necessarily a faithful account of its internal process.

Show how information, integration and human work affect the result. Avoid demonstrations that hide the review required to make an output usable.

Examine the investment case

Translate the proposed use into an operating consequence. Which outcome changes, what is the baseline and how does the benefit reach the business?

Include implementation, checking, support and capacity use. Directors should be able to distinguish learning investments, local productivity and sustained operating value.

Understand authority and control

For an agent, what may it read, write, send or decide? Which actions need meaningful review? What is prohibited and how is that boundary enforced?

For consequential uses, ask where specialist security, privacy and legal assessment belongs.

Practise on a decision paper

Use a clearly labelled example to examine the recommendation, alternatives, assumptions, controls and owner. Ask directors what information would change their decision.

A useful exercise surfaces uncertainty rather than presenting a single obvious answer.

Leave with practical questions

What have we authorised? Who owns the outcome? What evidence supports expansion? What remains unresolved? Can the organisation stop or recover the use?

Follow the session with management actions linked to the real portfolio. Education can improve oversight, but it does not certify governance or replace professional advice.

Explore a board briefing or use the board paper guide.