Skip to content
Schedule a consult
Schedule a consult

SECOND READING · 21 November 2026

Against AI Directors

Why a Board Is Not a College of Experts

I once worked with a board dealing with a serious cybersecurity incident. There was a lot of discussion, early on, about how to respond and when, but it was led almost entirely by the one cybersecurity expert on the board. Everyone else in the room fell silent. I think they made the right calls in the end, but this was not a group activity.

Today, boards are being told there is a new seat they must fill. This isn’t new. After Enron, it was the financial expert. A decade ago, it was the cybersecurity director. Now it is the AI expert, and you can see the pressure is real. In Corporate Board Member’s 2026 survey of US public-company directors, AI expertise ranked fourth among the attributes boards want in their next appointee. Only industry, financial and C-suite experience ranked higher. Among Fortune 100 companies, the share mentioning AI in director qualifications jumped from 26 to 44 percent in a single year, according to EY.

This instinct is understandable. AI is reshaping business models, and a director who cannot ask intelligent questions about it isn’t doing their job. But the idea that every board needs its own AI director rests on a mistaken view of what a board is.

A board is not a ‘College of Experts,’ each the master of a separate domain. It is a deliberative body. Real board-led value lies less in what any one director knows than in the judgment the group reaches together. Does it test management’s assumptions, weigh competing risks, and say no when it should? This sort of thing takes generalists: people with breadth, a strategic sense, and the confidence to ask the obvious question in a room full of experienced people.

A designated expert undermines this in two ways. The first is about knowledge. It is very hard to contradict an expert on a subject you know little about. When the AI director says a model’s risks are manageable, what does the former finance director beside her say? The second is about behavior. It is too easy to sit back. Once one person “owns” a topic, the others feel free to stay quiet, and over time they stop learning. Debevoise & Plimpton made a similar point this spring, warning that directors may “defer excessively to an AI expert, reducing the level of constructive challenge and debate.” This was the lesson of my cyber-incident board.

A board is not a college of experts. It is a deliberative body.

Yet regulation and market practice have both drifted towards the expert model. From 2021, Japan’s Corporate Governance Code asked listed companies to disclose the combination of skills their directors hold, typically as a skills matrix. Australia’s ASX principles have long recommended one too. The US has no matrix rule, but since 2009 the SEC has required companies to explain, director by director, the experience and skills that qualify each nominee. The market has gone further on its own: 80 percent of S&P 500 boards now publish a skills matrix, up from 38 percent in 2020, according to Spencer Stuart. These grids have made boards more transparent, and they help nomination committees spot real gaps. But they also reward boards for collecting checks in boxes: one for finance, one for cyber, and now one for AI. A matrix records what directors know. It says nothing about how they think, whether they challenge, or whether they change their minds when evidence changes.

I see some signs of a correction. The draft fifth edition of the ASX principles, released in July, drops the recommended disclosure of a skills matrix. Instead it asks boards to decide what collective capabilities they need and to explain how they assess them. Philip Lowe, who chairs the advisory group behind the draft, put it simply: if a matrix is useful to you, publish one, but it is not the only way. Japan has stepped back too. Its revised Code, finalized in July, still asks boards to identify the skills their strategy needs and explain how directors’ skills combine, but the matrix itself now sits in non-binding guidance, as one way of doing so. The US had a version of this debate in 2023, when the SEC removed a proposed requirement to disclose board-level cybersecurity expertise from its final cyber rules. The UK Code asks boards for a combination of skills, experience and knowledge. It pairs that with a principle that no individual or small group should dominate the board’s decision-making. I think this pairing is the right instinct.

And the strongest objection deserves a hearing. Some companies are AI developers, chipmakers, or businesses whose product is the model. For these companies, deep technical knowledge in the boardroom is essential, and a board of generalists will be outmatched by management. This is fair. But even here, the expert should add to the qualities every director needs, not replace them. Nor is “cultural fit” the answer to the opposite problem, though I see many boards treat it that way. In PwC’s 2026 survey, 81 percent of directors rated fit with the board’s culture as very important in a new candidate, against 27 percent for specialist expertise. A board of comfortable generalists that never challenges is no better than a panel of experts nobody challenges.

What should boards do instead? I think there are five ideas worth considering:

  1. Make AI literacy every director’s job, not one director’s portfolio. That means education, time with the people building and deploying the systems, and thinking about questions directors should be able to ask. The AI Board Governance Principles I co-authored with KPMG and INSEAD take the same view: every director needs sufficient AI fluency, but not every board needs a designated AI expert.
  2. Buy expertise in a form the whole board can challenge. Use outside advisers and management briefings that every director questions. In the US, Delaware law already protects directors who rely in good faith on experts chosen with reasonable care.
  3. If you do appoint a technical director, make sure the Chair is a guardrail against the dangers outlined here: the Chair should make sure that the technical director’s views are tested, not deferred to.
  4. Use the board evaluation to ask a harder question. Not whether skills appear on the matrix, but whether directors challenge each other, and on which subjects they go quiet.
  5. Regulators and investors should ask boards to explain how they assess their collective capability and judgment, rather than counting ticks. This is easier said than done.

We have become very good at measuring what directors know, but much less good at assessing how they think.


Dan Konigsburg is the founder of Konigsburg & Co., an independent adviser on board effectiveness, partnership governance and ownership. He chaired the OECD BIAC Committee on Corporate Governance from 2016 to 2025. Second Reading takes a second, harder look at governance ideas the market has accepted. Views are his own.

Notes and sources

  1. Corporate Board Member and Diligent Institute, What Directors Think 2026.
  2. EY Center for Board Matters, Cyber and AI oversight disclosures: what companies shared in 2025 (Fortune 100).
  3. Debevoise & Plimpton, Board Oversight of AI: Do Boards Need AI Experts?, April 2026.
  4. Spencer Stuart, 2026 U.S. Spencer Stuart Board Index; 2024 U.S. Board Index (2020 figure).
  5. SEC Regulation S-K, Item 401(e); see Compliance and Disclosure Interpretations, Section 116.
  6. Australian Institute of Company Directors, ASX releases draft 5th edition Corporate Governance Principles, July 2026; interview with Dr Philip Lowe.
  7. Tokyo Stock Exchange, The 2026 Revision of Japan’s Corporate Governance Code.
  8. Davis Polk, SEC adopts cybersecurity disclosure mandates for public companies, July 2023.
  9. Financial Reporting Council, UK Corporate Governance Code 2024, Principles G and K.
  10. PwC, 2026 Annual Corporate Directors Survey.

All essays