Better controls clear a path for AI in finance

As financial processes increasingly rely on AI, establishing robust controls becomes critical. Executives face pressure to adopt AI faster, yet many organizations lack data quality and controls necessary to trust AI output. Workiva Inc. aims to address this gap by applying established reporting safeguards to AI-assisted processes.

However, research suggests corporate confidence in AI-assisted financial reporting has moved ahead of operational readiness. Steve Soter, Workiva's vice president and industry principal, noted that 84% of executives are willing to trust AI to generate an annual report without human review – a figure that surprised him. The risk extends beyond minor errors, as without traceable information and documented review, AI could turn individual mistakes into broader control failures.

Established financial controls remain relevant when AI handles more work, as Soter explained. "To me, I think it’s maybe a different flavor of the same risk," he said. "When I think about it, back to the days when I was a controller, it was really important for me to know where the data was coming from, who touched it, what happened to it, how did it get reviewed and approved?"

AI can speed up established reporting processes, but speed alone is insufficient without a reliable underlying workflow. Financial teams still require governed data and documented approvals. Otherwise, automation could disseminate errors more rapidly and complicate tracing the source of issues. Human oversight remains crucial, especially when AI handles information for boards or external audiences.

While AI may generate the material, final approval remains with executives. This underscores the importance of review as a core part of the reporting process rather than a temporary measure while the technology evolves. "An AI tool isn’t signing off on the financial statements; a human is," Soter emphasized. Data quality presents another challenge.

Only 11% of executives believe their data is sufficient for AI use, indicating many organizations are automating processes before repairing their information foundations. While AI may help improve data quality, unreliable inputs will continue producing questionable results. Soter noted, "It makes you wonder, how bad was the data before we were even having this AI conversation?"

"To me, that just underscores, honestly, the opportunity for AI, because I think AI actually has a role in potentially helping to clean that up, like maybe boosting that 11%, but AI is only as good as the data that it is using."