How a board can assess AI competence in a CEO candidate

A board can easily ask a candidate for a view on AI. Most senior executives can now deliver one. The harder task is to distinguish a leader who has formed and acted on a considered view from one who has absorbed the market's language. That requires assessment around real decisions, with consistent questions asked of each candidate and evidence sought beyond the interview.

The aim is not to turn directors into technical examiners. It is to understand whether the candidate can govern a consequential change in products, operations, people and risk. A CEO does not need to train a model. They do need to recognise where technical claims affect strategic choices and when independent expertise is required.

Ask for a decision narrative

Invite the candidate to describe one AI-related decision in which they had meaningful responsibility. What problem was the organisation trying to solve? What alternatives were considered? Which data or capability constraints mattered? Who challenged the proposal? What was approved, changed or stopped? What happened after the initial launch? This sequence is more revealing than a question about favourite tools.

The board should probe the candidate's actual role. Did they choose the investment, set risk appetite and change incentives, or did they receive updates from a team that did the work? Both experiences can be valuable, but they are not equivalent. A candidate should be comfortable naming the technical and operational people whose expertise shaped the outcome.

Test the economics

Ask how the candidate distinguished a successful demonstration from durable value. Which costs did they include: data preparation, integration, oversight, change management and ongoing operation? Which benefits were measured, and over what period? What would have caused them to stop? A leader who can explain the economics of a small, unsuccessful initiative may demonstrate better judgement than one who quotes a large theoretical productivity opportunity.

The questions should match the client's strategic context. A regulated business may care most about controlled use and trust. A service business may care about quality, client experience and how professional work changes. A product company may need a deeper account of defensibility and speed. Generic AI enthusiasm is difficult to interpret without a specific business problem.

Test governance under pressure

Present a hypothetical but realistic tension: a business unit wants to deploy a promising AI tool before data controls are complete; a competitor has announced a similar product; the expected benefit is material. Ask what the candidate would require before a decision, who would be involved and how they would handle disagreement. There is no single script to memorise. The board is looking for proportionality, clarity of accountability and the ability to act without pretending uncertainty has disappeared.

Look for learning capacity

Because the technology and market are moving, the candidate's present knowledge has a short shelf life. Ask how they update their view. Whose expertise do they seek? When did new evidence change their mind? How do they prevent an expert team from becoming a black box the CEO cannot challenge? These questions assess the habit of learning rather than the possession of fashionable vocabulary.

A fair assessment should end with a written comparison against the mandate: strategic choice, business economics, governance, talent leadership and learning. It should distinguish observed behaviour from self-report and identify remaining questions for references or later conversations. The board then has a basis to judge AI competence as part of overall CEO readiness, rather than as a separate buzzword on the scorecard.

One practical exercise is to give all finalists the same short strategic scenario. Describe a proposed AI investment, an uncertain economic case, a concern from the risk function and a competitor announcement. Ask each candidate for the questions they would put to management, the decision they could make now and the evidence they would require for a larger commitment. The assessment should reward a coherent process, not agreement with a preferred answer. It should also examine whether the candidate can explain the trade-off to a board and to employees who will live with the result.

References can then test what the interview cannot: whether the candidate listened to expertise, owned difficult decisions and changed course when evidence warranted it. The reference questions should be specific to an authorised, relevant example and respect confidentiality. A polished answer in the interview is one piece of evidence; consistent conduct under pressure is the stronger signal.