Should we appoint a director with AI expertise?

Two-thirds of boards say they do not know enough about AI, and the usual response is to look for a director who does. Deloitte's 2025 survey of 695 board members and executives across 56 countries found 66 per cent describing their boards as having 'limited to no knowledge or experience' with AI, 31 per cent saying it is not on the board agenda, and 40 per cent saying AI has made them think differently about the board's makeup. The question is what the board is trying to become able to do. If the answer is to judge management's allocation of decisions between people and machines, the specialist director is one option among several and not obviously the best. If the answer is to understand the technology, the board has confused its job with the chief technology officer's.

What a board has to be able to do#

Four things, and none of them is technical. Know which decisions management has allocated to machines, to inform, recommend or execute, and which it has kept human. Know who owns each, by name. Know what the organisation is still capable of doing unaided, and how it would find out if that capability had gone. And know how often the humans reviewing machine output actually reach a different answer, because a review that never disagrees is a signature rather than a review. A board that can insist on those four documents, read them and ask what changed since last quarter is doing its job on AI. A board that cannot has delegated more than it intended, whatever its members know about transformers.

What one specialist director supplies, and what happens next#

A director with real AI experience supplies two things: the ability to tell when management's account of a system is technically implausible, and the standing to say so. Both are valuable. The risk is what the rest of the board does with them. In most boards a single specialist becomes the person AI questions are routed to, the other directors stop forming views, and the board has reproduced at its own level the failure this site documents at executive level: judgement delegated to whoever understands the tool. Banking's model risk regime asks for effective challenge from people with independence, standing and expertise, and is careful to say that expertise is the part that has to be maintained by doing the work. One person cannot be the board's expertise; they can only be its excuse for not building any.

The alternatives, and when each fits#

A standing adviser to the board, without a vote, gives the challenge without the delegation, because the adviser cannot be the one who decides. An annual external review of the allocation, meaning the list of decisions, owners, capability floor and disagreement rate, gives the board an independent reading of the documents rather than a person to defer to; the case for independent assurance is at how a board knows management's claims about AI are true. And the whole board using the tools, on real work, for a month, does more for the quality of its questions than any appointment, because the three failure modes of oversight, how confident the output sounds when wrong, how much it does unasked and how quickly one stops checking, are learned by experience and not by briefing. The specialist director fits when the company's product is the AI, when the board will be asked to approve technical bets it cannot otherwise read, or when the sector regulator expects it. For most boards, the sequence is the four documents first, the adviser second, the appointment if the first two show a gap nobody at the table can close.

What nobody has measured#

No study relates board AI expertise to outcomes, and Deloitte's figures are self-descriptions from a panel that its authors do not claim is representative. 'Limited to no knowledge' is what a careful director says about a subject that is changing monthly, and may overstate the gap. The argument here is about what a board's job on AI consists of, which is checkable against the regulation and against where AI programmes fail, rather than about whether appointing anybody helps, which is untested.

What the board asks for at its next meeting#

The allocation, in writing. The owner of each decision on it. The statement of what the organisation keeps practising unaided. The disagreement rate, or the admission that nobody measures it. Then the board knows whether its gap is expertise or documents, and in the experience behind this site it is nearly always documents. The questions in full are at what should a board ask about AI, and the oversight that results at what board oversight of AI looks like.

Key sources

• Deloitte Global Boardroom Program (2025). Governance of AI: A critical imperative for today's boards, 2nd edition. Graded entry.
• Singla, A. et al. (2025). The state of AI. McKinsey. Graded entry.
• Vaccaro, M., Almaatouq, A. and Malone, T. (2024). When combinations of humans and AI are useful. Graded entry.
• Bainbridge, L. (1983). Ironies of automation. Graded entry.

Related SuperSkills research#

The board's job on AI is set out at what board oversight of AI looks like and what should a board ask about AI. On whether the job needs technical people, do you need to be technical to lead AI. On the difference between the machinery and the decisions, AI governance versus AI leadership. The advisory seat is described at board advisory.

About this research#

Rahim Hirji is the author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. He has run, grown, bought and advised businesses with AI in them. Findings are attributed to the studies and statements that produced them and kept separate from the interpretation. This is a living reference, reviewed and updated as significant new evidence appears.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.