top of page
Search

How I think about AI at the audit committee level

AI is showing up on audit committee agendas now. Not as a deep topic. As a recurring question. How is the company using AI in the finance function, and what risks does it create.

The question is reasonable. Most of the answers I hear are not.


Finance teams describe their AI usage to audit committees the way they describe any new tool. What it is, what it does, how it fits the stack. That framing is incomplete. Audit committees are not asking about tools. They are asking about exposure.

The version of the answer that matters has three parts.


Part one is where AI currently sits in the finance workflow. Not a list of tools. A specific map of where in the process AI is producing output, who reviews that output before it leaves the finance function, and what the review catches when it catches something. Audit committees can assess controls. They cannot assess tools. The framing has to be in their language.

Part two is the data question. What data is going into the AI systems the finance team uses. Is any of it confidential or regulated. Are the tools running on a tenant that isolates the company's data from model training. If the answer to any of these is uncertain, that is the first thing the audit committee wants to know.


Part three is the disclosure question. If AI output ever materially influences a number, a disclosure, or a filing, how would the company know. This is the harder question. Most finance teams cannot answer it yet. The honest answer is that AI usage is diffuse enough now that a robust audit trail does not exist. That answer is uncomfortable. It is also more credible than a confident story that would not survive five minutes of inspection.


The audit committees that ask these questions well are signaling where governance is headed. The finance leaders who answer them honestly are building the credibility they will need when the governance expectations catch up.


When did your audit committee last get a real answer on how AI is being used in the finance function, and what happens if something goes wrong?

 
 
 

Recent Posts

See All
What AI cannot do in an audit

AI is already changing how audits run. Document review, sampling, risk identification, and analytical procedures are all getting faster. The direction of travel is clear. That is not the interesting q

 
 
 
Why finance teams should write their own prompts

A lot of finance teams are waiting for prompt libraries to arrive from vendors, consultants, or enablement teams. That is a reasonable way to save time. It is also the wrong way to build AI capability

 
 
 

Comments


bottom of page