Auditors want CFOs to show their AI work

As more finance teams pull in AI to help them draft accounting policies and memos, close the books, and other financial reporting tasks, more auditors are left wondering about the accuracy of their clients’ data and workpapers.

On Oct. 22 at Campfire’s Basecamp summit in San Francisco, a panel of finance leaders discussed AI-enabled auditing and the importance of leaving a work trail—not only for the company but for its accountants—and how to make it transparent enough to satisfy the peskiest auditors.

Panelist and controller Emily Miller of nutrition counseling platform Nourish said her company’s goal “is to try to push as many AI-enabled workflows as we can,” and as a result “we spend our time making sure that we actually create those breadcrumbs for our auditors to follow.”

The audit trail is about “just making sure that all the back and forth is there; show your work with the AI,” Miller said. “You can export your conversations with Claude and OpenAI and leave it as evidence…have that available so that your auditor feels comfortable,” she said.

“At the end of the day, our auditor is going to keep our feet to the fire to make sure that we have someone actually reviewing this,” Miller added. “An AI agent is a thing that is relatively deterministic. It’s not there to give you a reasonability assessment; that’s your job as the human: ‘Is that actually reasonable at all?’”

No experts. From the perspective of the auditor, panelist and Deloitte partner Jeff Jenkins reminded the crowd that auditors are learning about AI at the same time as finance teams and CFOs. “This is an emerging technology for us as well on the audit side,” Jenkins said. “Sometimes it can feel like, well, ‘the auditors must have some game plan that’s all figured out. They know 100% what to do here.’”

With that in mind, CFOs should be cognizant of how their auditor uses AI versus how the company uses AI, “particularly around financial reporting, [the] accounting close, and how that might impact what the auditor can do,” he said.

“If you come to the auditor and say, ‘We want you to come audit all of our close processes and do this year-end audit’…and then the auditor asks, ‘Well, tell me about your close process,’ and you say, ‘Well, we’ve got 100 Excel spreadsheets, then there’s this guy who has all the fixed assets memorized in his head, and he inputs that at the end of the quarter,’ that’s a very different scenario than saying, ‘We’ve got an AI-native ERP, and we’ve got some agentic workflows, and that all works together to close our books and create our financial statements.’ Those are very different things,” Jenkins added.

“IT and tech maturity matters a lot,” he said, “inasmuch as how much the auditor can rely on that and get comfortable with what you’re doing, versus maybe a more maturing process that’s a little more manual, spreadsheet-based.”

In some cases, when AI is involved in an audit trail, Deloitte deploys an AI specialist, Jenkins said. “If someone hands me code, I can’t read code.” The AI specialist “would come read it and help us understand, ‘OK, the company’s saying this does X. Is that actually what it’s doing?’ And there’s a whole slew of things that come with that around, how’s that code controlled from an access perspective?”

Despite the excitement of “all the AI [auditing] capabilities,” the human elements must be prioritized in areas of governance, risk, and controls, Sameer Syed, panelist and chief accounting officer at AI procurement platform Zip, said.

“There’s some really impressive technology, but if [companies] don’t think about that controls layer, that governance layer, it doesn’t matter. You’ve opened yourself up to risk, fraud, etc., and that’s just not acceptable,” he added.

Lower fees ahead? Half-jokingly, Syed asked Deloitte’s Jenkins if AI-enabled auditing would reduce the fees that companies pay for audit services. Jenkins gave the laughing crowd the good news: Fees likely will come down, but not in the short term.

“There is this investment of, what does an audit with AI in it look like, and how do you build that on both sides? That’s not going to decrease the cost on day one, because it’s an investment of time, resources, and effort,” he said.

“I do think in the longer tail you could see a model where you hook—you being the auditor—a proprietary tool onto an ERP, it ingests the data, and then some high percentage of transactions—maybe it’s 70%, maybe it’s 80%—are audited routinely,” Jenkins said. “Because it’s just GAAP, which is a set of rules applied to data, and 80% of that is blocking and tackling, then the judgment is handled by humans, and maybe some agents take a first cut at that. I could see the future of audit going to that eventually…But I do think what will happen in the near and medium term is, we’ll see probably team sizes shrink; we’ll see more AI deployed from the auditor to do things faster, smarter, better.”