AI governance has moved out of the IT function. The question for boards is no longer whether to engage with artificial intelligence, but whether they can afford the cost of continuing to delegate it.
Current data shows AI oversight remains a minority practice in major corporate boards. Glass Lewis research, published via the Harvard Law School Forum on Corporate Governance in March 2026, found that just 54% of S&P 100 companies disclose board-level AI oversight, and only 28% disclose both oversight and a formal AI policy. A wider Thomson Reuters Foundation study from February 2026, covering 1,000 companies across 13 sectors, found that only 38% of companies in the Americas have published an AI policy, despite the US being the world's leading hub for AI innovation.
Beyond disclosure, the deeper problem is competence. The Deloitte Global Boardroom Program surveyed nearly 700 board members and senior executives across 56 countries and found that 66% report their boards have "limited to no knowledge or experience" with AI, while nearly one in three say AI does not appear on their agendas at all. These are the boards of some of the largest and most consequential organisations in the world.
This is a governance failure. It is also, increasingly, a performance and liability issue.
The market is not waiting for boards to catch up. In the 2024 proxy season, the number of AI-related shareholder proposals more than quadrupled year-over-year, reflecting a significant escalation in investor attention. The proposals are specific and substantive: third-party evaluation reports, workforce impact assessments, transparency on AI decision-making in areas ranging from hiring to credit underwriting.
The proponents include labour unions such as the AFL-CIO, faith-based investors, and ESG-focused shareholders including Trillium Asset Management and Arjuna Capital. This is a coalition with reach and persistence. These groups have demonstrated, across other governance issues, that they pursue long campaigns and they win.
The message to boards is clear. If AI governance cannot be evidenced in proxy disclosures, in board minutes, in committee structures, shareholder resolutions will force the issue. Regulatory attention in both the US and UK is moving in the same direction. The window for voluntary, considered governance is narrowing.
Governance arguments alone rarely move boards at pace, however performance arguments do.
A 2025 study from MIT found that organisations with digitally and AI-savvy boards outperform their peers by 10.9 percentage points in return on equity.
Boards with genuine AI literacy are better positioned to ask the right questions of management: whether AI investments are generating measurable returns, where AI introduces operational or reputational risk, how AI-related capital allocation decisions compare against alternative uses of the same budget. These are exactly the questions that non-executive directors are supposed to ask. Without the foundational understanding to ask them well, oversight becomes nominal.
Yet an EY review of proxy statements filed between January and November 2025 found that only 12% of Fortune 100 companies disclosed that board members had received any AI education or training. The majority of boards are being asked to govern technology they have not been equipped to understand. The Thomson Reuters Foundation identified a related problem further down the organisation: of companies that do have AI policies in place, only 41% make those policies accessible to employees or require acknowledgement of them. Governance on paper is not governance in practice.
The point is not that boards need to become technical. Chief Information Officers and Chief Technology Officers exist for precisely that reason. The point is that AI now sits alongside M&A, talent strategy, and regulatory risk as a category that requires genuine board-level strategic engagement, not periodic reporting from management.
In practice, that means several things. Boards should be reviewing AI strategy as a standing agenda item, with the same rhythm and rigour applied to financial performance. Audit or risk committees need remits that explicitly include AI governance, covering model risk, data quality, third-party AI dependencies, and regulatory exposure. Boards should be asking management to demonstrate where AI is creating measurable value and where it is creating concentration risk.
It also means investing in board-level capability. Structured education programmes, access to independent technical advisers, and where appropriate, recruitment of directors with relevant expertise are all legitimate tools. The current disclosure rate on AI training is not a benchmark to aspire to; it is a warning of how far most boards still have to travel.
Boards that treat AI as an IT function will continue to make decisions about AI without understanding its strategic implications. The organisations with boards that have done the work will outperform. The organisations that have not will face growing pressure from shareholders, regulators, and the market itself to explain why.
The governance expectation has already shifted. The performance evidence is already in. For senior leaders who have been waiting for the right moment to put AI on the board agenda in a serious way, this is it.