If generative AI is a potential game-changer for finance, agentic AI, which acts proactively rather than reactively, is even more so. But most CFOs say they feel pressed to deploy agents more quickly than is practicable.
That’s according to a June survey of 1,505 CFOs and other senior finance leaders in the United States, United Kingdom, Australia and India who had deployed, piloted or actively evaluated AI agents in the previous 12 months.
The research, performed by Censuswide, was commissioned by Avalara, an interested party because it’s a provider of agentic tax and compliance software. However, the findings are eye-opening nonetheless.
An overwhelming majority of those polled, 92%, said they’re under significant (50%) or moderate (42%) pressure to demonstrate ROI from AI agents, and 71% said the pressure is entirely or mostly around deployment speed.
The kicker: Only 7% of the finance leaders said their organizations value governance over speed when it comes to agentic AI. Among U.S. respondents, the proportion was 4%.
Three in 10 respondents said their organizations had not updated internal controls within the preceding year to reflect AI agents taking or recommending actions, and 44% said they are only somewhat confident they could explain an AI agent’s actions to an auditor or regulator.
According to Avalara, the findings reveal a finance function caught between executive pressure to accelerate agentic AI adoption and the operational reality that agents need to be managed with care, particularly in tax and compliance, where decisions must withstand regulatory scrutiny.
“Speed without accountability creates new forms of risk, and speed without rethinking workflows limits ROI,” said Avalara CEO Hugo Sarrazin in the report.
Not only don’t CFOs feel confident they understand agentic AI, 76% of respondents indicated that their organization lacks dedicated in-house finance expertise to understand how agents work.
About a quarter of the survey participants (23%) said accountability for a significant AI error would be unclear or sit with no one.
Meanwhile, although 38% of respondents said they are deriving ROI “at scale” from AI agents, 50% said such return is limited and 12% said there is not yet any clear ROI.
