What boards should ask before they trust an AI-assisted audit

AI-assisted assurance tools can review more evidence, faster, than any human team. That's the pitch, and it's largely true. What gets lost in the pitch is that the board's job doesn't shrink when the review is automated, it changes shape.

The right question isn't 'does the AI catch more issues.' It's 'what does the AI decide not to flag, and who reviews that decision.' Every automated assurance tool makes judgment calls about materiality. Boards need to see those calls, not just the output.

We built Nicole, our own audit AI, around that principle: the model surfaces evidence and discrepancies, but a human assurance lead still owns the sign-off. Speed without a visible decision trail isn't assurance, it's just a faster unverified claim.