At a recent session I led for current and aspiring directors, a trustee of an independent school described what happened when his board decided it was time to take up AI.
The board expected to set direction. Instead it found that the school was already well along. A division head had partnered with the chief information officer, adopted a framework based on outside research, and launched initiatives covering what students should learn about AI, how it could save teachers administrative time, and how it changes what comes after graduation. The board formed a task force, but its first job was to get smart about what the school was already doing. In his words, the board was catching up.
That story is more common than most boards would admit, and it isn’t a failure. Management should be moving faster than the board on operating questions. The problem starts when the board can’t tell whether what management is doing is the right thing, because it doesn’t know enough to ask.
Directors know they are behind. In PwC’s 2026 Annual Corporate Directors Survey, 71% of directors said their boards need to strengthen their AI skills to provide more effective oversight. And 82% rated the information they receive linking AI outcomes, risks and business performance as fair, poor or not provided at all (PwC). That second number is the real problem. A board can’t oversee what it can’t see.
Part of the gap is structural. AI moves at operating speed, and boards meet four to six times a year. A quarterly update from the CIO is out of date before the next one arrives. Another director in the same session, a product and technology leader who works with these tools every day, made the point that no framework can keep pace with how quickly AI is changing. Her view was that effective oversight requires some current, hands-on familiarity on the board itself. Without it, directors are left evaluating management’s AI strategy through management’s own description of it.
Some boards are responding with structure. AI oversight often starts as an agenda item, moves to the audit committee and eventually gets a committee of its own, the way audit and compensation committees evolved before it. Regeneron is a recent example. Its 2026 proxy statement describes a new standing Digital Technology Committee, established in April 2026 and made up entirely of independent directors. Its remit includes digital technology strategy, including AI and machine learning, the risks and controls around how the company uses that technology, digital transformation priorities and cybersecurity, which the audit committee oversaw before.
Most boards won’t create a new committee, and many shouldn’t. But every board can ask better questions, and four cover most of the ground.
Where is AI actually being used? Not only the projects management sponsored, but also the tools employees adopted on their own. Most boards don’t have a clear answer, and many managements don’t either.
What gets funded, and what gets stopped? A long list of pilots can look like progress while spreading people and money thin. The board should understand how management decides which initiatives move forward.
How do we know it’s working? Ask what value has been realized, how it was measured and who measured it. “We’re experimenting” is a reasonable answer for a quarter or two. It is not a strategy.
What could go wrong, and who owns it? Data privacy, vendor and model risk, errors in regulated work and reputational exposure all need an owner in management and a place on the board’s calendar. Frameworks like the NIST AI Risk Management Framework are short and free, and a director who has read the overview will ask sharper questions than one who hasn’t.
Asking those questions well depends on the currency the board was missing in the first place. A few practices help close the gap. Ask for a standing briefing from whoever owns AI in management, with time for discussion rather than a status report. Make sure at least one director has recent operating experience with these tools, and use the board’s skills matrix to find out whether you do. Stay current between meetings instead of relying on the quarterly update.
And use the tools. Directors who have used AI themselves on real work ask better questions about it than those who have only read about it. The board’s own preparation is a natural place to start, provided it happens in a tool the organization has approved rather than a personal account. Many boards haven’t been given one. Deloitte’s Center for Board Effectiveness reported this summer that nearly half of public companies have not formally enabled or standardized AI for board activities (Deloitte). That is a large part of why we built Aureclar: it gives directors a governed place to use AI on their own board work, along with a news feed that keeps them current on their organizations and sectors between meetings.
Catching up is normal. The board’s job is to close the gap quickly enough that it can tell management something management doesn’t already know.
Josh Schiffman is the founder and CEO of Aureclar, AI for the boardroom that helps directors prepare and exercise better judgment.