Dive Brief:

• Ninety-two percent of CFOs and top finance executives feel pressure to show that investment in artificial intelligence yields a decent return, and only 7% say their organization focuses more on AI governance than speed of adoption, Avalara discovered through a survey.
• Forty-four percent of respondents are only somewhat confident they could explain the actions of an AI agent to an auditor or regulator, and 76% say their organizations lack the in-house expertise sufficient to understand how their AI operates, Avalara, a provider of AI tax compliance software, said Tuesday.
• “Nearly every leader we surveyed feels personal pressure to prove these investments are delivering,” Avalara Executive Vice President Jayme Fishman said in a statement. “Meanwhile, the controls, expertise and accountability meant to sit underneath them are still being built.”

Dive Insight:

Worldwide spending on AI will surge 47% this year to $2.6 trillion from $1.76 trillion in 2025, according to Gartner, exceeding the gross domestic product of both Canada and Australia. By 2030, AI investment will rise 120% more to $5.62 trillion.

The rush to adopt AI thrusts CFOs into a dilemma: They must fulfill their role as the main line of defense against wasteful spending while facing institutional, investor and competitive pressure to fund AI projects that, on their face, promise big returns.

“What stood out to me is the tension finance leaders are managing every day,” Avalara CEO Hugo Sarrazin said in a statement. “They're being asked to move quickly, prove value and modernize critical processes while also protecting the controls and governance their businesses depend on.”

In one sign that AI oversight lags adoption, 30% of survey respondents said their company has not updated internal controls within the last year, Avalara found in its survey of 1,505 CFOs and senior finance executives.

Half of survey respondents said their AI agents have yielded only limited measurable ROI, Avalara said.

Accountability for AI errors is also spotty, with 23% of respondents saying responsibility for a mishap would fall to no one or would be unclear, Avalara said.

Also, 46% of those surveyed said their AI incident response plans are untested or still in development, and only 28% require documented audit logs showing how their AI agent reaches its decisions, according to Avalara.

Companies should build and operate AI agents using verified, domain specific data and ensure transparency in decision-making, Sarrazin said.

Clear accountability and governance should undergird AI agents before, rather than after, launch, and companies adopting AI should choose partners with expertise in finance and compliance, he said.

Avalara’s survey yielded responses in June from CFOs and other senior finance leaders at companies with annual revenue of at least $10 million in the U.S., U.K., Australia and India. Respondents deployed, piloted or evaluated AI agents in their organizations for at least the previous 12 months.