Avalara, Inc., a leader in agentic AI for global tax and compliance, has released new research showing that while finance teams are under pressure to deploy AI agents as quickly as possible, governance, accountability, and internal controls are struggling to keep pace.

The report, Agents of Change: How the Race to Deploy AI Agents is Outrunning Financial Governance, surveyed more than 1,500 CFOs and senior finance leaders in the United States, United Kingdom, India, and Australia who had deployed, piloted, or actively evaluated AI agents for financial processes over the past year.

Executives Face Pressure to Show Results

The survey shows that 92% of respondents feel moderate or significant pressure in their careers to demonstrate that investments in AI agents are generating a return on investment. Despite this pressure, half say their AI agent initiatives have delivered only limited measurable ROI.

An interesting finding is that CFOs are facing pressure not only to deliver results, but also to accelerate the pace of digital transformation. According to the survey, 71% say the pressure to deploy AI is primarily focused on speed.

AI and Governance Are Not Moving at the Same Pace

Although the survey shows that AI is a unanimous topic among executives, innovation appears to be outpacing governance. Only 7% of respondents say their organizations prioritize governance over speed when measuring AI projects.

The research also shows that 30% of companies have not updated their internal controls over the past year to reflect situations in which AI agents have begun executing or recommending actions. Another 44% say they have only moderate confidence that they could explain an AI agent’s actions to an auditor or regulatory authority.

“Financial leaders are being pressured to accelerate AI adoption, but agent governance requires a new combination of business, artificial intelligence, information technology, and data governance expertise,” said Frank Cirone, Vice President of Commercial Strategy at Snowflake, a cloud data platform company.

“As AI agents gain access to financial and compliance workflows, organizations need to know what information these agents can access, what actions they can perform, and when human approval is mandatory. This level of control needs to be built into the solution architecture from the outset, rather than added later,” he added.

Another important issue is the depth of AI expertise within companies themselves. The survey reveals that more than three-quarters (76%) of the leaders Avalara spoke to said they currently lack dedicated expertise within their finance departments to understand how their AI agents work.

“You can’t give a tax team an AI tool and expect transformation, just as you can’t ask AI engineers to solve complex tax problems on their own. The strongest results come when tax professionals and AI specialists work together to build solutions that are accurate, practical, and scalable,” said James Erwin, Head of Indirect Tax.

This willingness to implement such a powerful new technology despite the lack of necessary expertise raises numerous concerns. One financial services leader in India summarized the risk:

“My biggest concern is that AI agents could make inaccurate or non-compliant decisions without timely detection, leading to financial, regulatory, or reputational risks for the organization.”

Who Is Responsible When AI Fails?

The research also highlights a fundamental question: who is responsible when an AI agent makes a significant error?

Nearly one in four respondents, or 23%, say responsibility for a serious error committed by an AI agent would be unclear or would not fall to any specific person or team. Another 16% believe that the executive who approved the AI investment would ultimately be held personally responsible.

Most respondents, however, say accountability for AI output would fall on the team or contractors responsible for implementing the technology, with 61% placing responsibility on the vendor or the individual or team that deploys or manages the agent.

Another factor increasing the risk is the lack of experience and maturity in AI incident response. While companies are beginning to establish procedures to address potential AI-related incidents, 46% said their AI incident response plans are either still under development or have never been tested, raising questions about how effectively they would perform in a real incident.

When asked what would most increase their confidence in expanding the use of AI agents, 27% said they would have greater confidence if the agents operated within existing systems of record.

Among the capabilities considered most valuable, 30% of respondents cited audit-ready documentation of all actions performed by AI agents. The same percentage highlighted continuous monitoring of regulatory changes, with real-time updates applied as regulations change.

“AI agents are beginning to operate within business processes that require trust, transparency, and governance from the outset,” said Jim Lundy, founder, CEO, and lead analyst at Aragon Research.
“As companies expand their use of agentic AI, the question is no longer whether the technology is capable of taking action, but whether organizations can understand, control, and explain those actions. In finance, where workflows are auditable and decisions have real business consequences, governance and explainability will become essential requirements for technology adoption.” Lundy added

About Avalara

Avalara is the agentic AI platform for global tax and compliance. For more than two decades, Avalara has built one of the most expansive libraries of tax content and integrations in the industry, processing more than 54 billion transactions annually and supporting millions of businesses worldwide.

The company’s purpose-built AI agents automate end-to-end compliance with greater precision, from tax calculations and return filings to exemption certificate management and beyond. For more information, visit Avalara.com.