Governance

AI oversight is a whole-board responsibility

The audit committee has real work to do on AI. But the questions that matter most, what AI is deciding and who answers for it, belong to the full board.

5 min

A dim office corridor beside a glass-walled server room

As companies adopt artificial intelligence, many boards have handed the topic to the audit committee. It is an understandable choice. Audit already looks at controls, data and risk, and it has the habit of asking for evidence.

But parking AI on one committee can leave the most important questions unasked. What the company uses AI for, and what it is prepared to let AI decide, are questions of strategy and risk appetite. Those belong to the whole board.

What the audit committee should cover

There is real work for audit. Where AI tools touch financial reporting, the committee should understand which processes are affected, what controls sit around them, and how management knows the outputs are reliable. It should also ask the external and internal auditors how they are approaching these areas.

  • Controls over the data that AI tools use in reporting and forecasting

  • Review and sign-off where AI output feeds into estimates or disclosures

  • Model risk: how models are validated, monitored and changed

  • Disclosure: what the company says publicly about its use of AI, and whether it is accurate

What the full board should own

Beyond reporting, AI is changing how decisions are made across the business: pricing, hiring, credit, customer service, product design. The board does not need to understand the technology in detail. It does need to know where AI is making or shaping decisions that matter, and who is accountable when it gets them wrong.

The board does not need to understand how the model works. It needs to know what the model is deciding, who answers for it, and what happens when it fails.

Three questions are a reasonable starting point for any board:

  • What is AI deciding? Which decisions, in which parts of the business, are made or materially shaped by AI tools today, and which are planned.

  • Who is accountable? For each significant use, which executive owns the outcome, including errors, complaints and regulatory questions.

  • What happens when it is wrong? How errors are detected, how quickly they are corrected, and how the board would hear about a serious failure.

Keep it on the agenda

AI should not be a one-off presentation. A short, regular report to the full board, alongside the risk register, keeps the topic in view as uses expand. Committees can then take the detailed work that fits their remit: audit on controls and reporting and, where one exists, a risk or technology committee on operational and cyber exposure.

Rules and guidance on AI are still developing, and they differ by industry and jurisdiction. That is another reason for the full board to stay involved. These are not only technical or compliance questions. They are questions about what kind of company the board is overseeing.

Selective engagements

Start with
a conversation.

No deck, no pitch. You talk, we listen, and together we find out whether we are the right fit.

Start a conversation

CLARUM

© 2026 Clarum Advisory LLC. All rights reserved.

CLARUM is the brand name under which Clarum Assurance LLC and Clarum Advisory LLC provide professional services. Clarum Assurance LLC is an independent CPA firm, and Clarum Advisory LLC is a business advisory firm. Professional services are provided by the applicable legal entity pursuant to the terms of each engagement.