The Condition

Survey data collected in 2025 from directors serving public-company boards shows that AI has moved materially onto board agendas. NACD reports that 62.33 percent of respondents set aside agenda time for full-board discussion of AI.

That is worth taking seriously on its own terms. Board attention is scarce. Moving AI into recurring board discussion signals that directors increasingly see it as consequential to strategy, workforce, competition, operations, and risk.

But attention answers a narrower question than visibility.

A board can discuss AI regularly and still lack the information it would need to exercise judgment over a consequential AI condition. Agenda presence tells us that the subject is being discussed. It does not, by itself, tell us what the board can actually see.

What the Evidence Shows

The useful evidence is not that boards are ignoring AI. The evidence points to something more specific: attention has moved faster than several of the formal practices that could convert attention into decision-relevant visibility.

NACD's 2025 data provides a useful comparison inside the same survey population. While 62.33 percent of respondents reported setting aside agenda time for full-board AI discussion, 36.30 percent reported adopting an AI governance framework, 27.40 percent reported incorporating AI oversight responsibilities into board committee charters, and 6.16 percent reported establishing metrics for management reporting.

Those figures do not prove that boards without those practices lack effective oversight. They do show that the most common reported activity is discussion, while several more formal oversight practices are less common.

A separate form of evidence comes from public disclosure analysis. McKinsey reported that, as of 2024, only 39 percent of Fortune 100 companies disclosed any form of board oversight of AI — whether through a committee, a director with AI expertise, or an ethics board.

Disclosure evidence has an important limitation: absence of public disclosure is not the same as absence of internal practice. A company may govern AI more actively than its filings reveal. But the evidence still matters because it provides a second, methodologically different view of the same general condition: board-level AI attention is increasingly visible, while formal oversight structure is less consistently visible.

The trend is also moving. Deloitte's second-edition global boardroom research found that the share of respondents saying AI was not on the board agenda declined from 45 percent in its prior survey to 31 percent in the later survey. The same research found that the share describing their boards as having limited to no knowledge or experience with AI declined from 79 percent to 66 percent.

That is not a picture of static deficiency. It is a picture of developing practice.

The institutional environment is developing too. In December 2025, the SEC Investor Advisory Committee approved a recommendation that included disclosure of board oversight mechanisms, if any, for AI deployment. The recommendation is advisory, not a rule. SEC Chairman Paul Atkins stated at the same meeting that he believes existing principles-based disclosure rules can address material AI information and cautioned against prescriptive disclosure requirements.

That disagreement is useful evidence in its own right. It shows that institutional attention is increasing without establishing a settled regulatory model for board AI oversight.

NIST's AI Risk Management Framework offers another point of reference. Its core is organized around four functions — Govern, Map, Measure, and Manage — and is intended for voluntary use. It does not establish a legal requirement for boards. It does, however, illustrate how AI risk governance is increasingly being described through structured responsibilities, information, assessment, and action rather than through attention alone.

The Distinction

Taken together, the evidence supports a narrow claim:

Among the populations examined, board attention to AI has increased faster than several of the formal governance practices that could help convert that attention into decision-relevant visibility.

That distinction matters.

A board can discuss AI at every meeting and still not have decision-relevant visibility if no one can identify where consequential AI information is supposed to land, if the information reaching the board is general rather than connected to what would actually matter, or if a material change cannot reach the relevant authority until the next scheduled meeting.

None of those conditions prove a governance failure.

They show why attention and visibility should not be treated as interchangeable.

The governing question is not simply whether AI appears on the agenda. It is whether the board can see enough of the consequential condition, at the right point of authority, to exercise judgment when the condition matters.

Where the Evidence Stops

The evidence does not establish that an attention-visibility gap causes AI failures.

It does not establish that adopting a formal AI governance framework produces better outcomes.

It does not establish that every board needs the same oversight structure.

It does not establish that the SEC currently requires a specific board AI oversight model.

And it does not establish that the presence of a committee, framework, metric, or recurring report is sufficient by itself to demonstrate meaningful oversight.

Those boundaries matter because formal structure and substantive visibility are not the same thing either.

A board may have a charter and still receive weak information. Another may operate through a broader enterprise-risk structure and still have a clear, functioning path for consequential AI conditions to reach the right authority.

The question is not whether an organization matches a universal model.

The question is whether the board can presently see what it would need to see for the decision in front of it.

What Would Actually Answer the Question

If attention and visibility are different conditions, a more useful board-level question is:

What would we need to see, from whom, and how quickly, to know whether we are positioned to govern a consequential AI condition rather than simply discuss AI as a subject?

That question does not have one universal answer.

But it creates a different way of examining board AI oversight.

It shifts the focus from how often AI appears on an agenda to whether consequential information has a clear destination, whether what reaches that destination connects to what actually matters, and whether material change can reach it in time.

Those are visibility questions.

And they can be asked without claiming that one board structure, one framework, or one set of metrics is right for every organization.

Close

Boards are paying more attention to AI. The evidence also shows that formal governance practices remain uneven and are still developing.

That is not a case for alarm. It is a reason to preserve the distinction.

A board that discusses AI is not, by that fact alone, a board that can see it.

Whether those are the same thing in any particular boardroom cannot be determined from the outside. It can only be examined where the information, authority, and decision actually meet.