Share with your CFO
Finance leaders are deploying agentic AI, autonomous software that initiates actions rather than waiting to be asked, faster than they can govern it, and the gap is starting to look dangerous. A Censuswide survey of 1,505 CFOs and senior finance leaders across four countries, commissioned by tax-compliance vendor Avalara, finds that 92% face significant or moderate pressure to demonstrate ROI from AI agents, while only 7% say their organizations prioritize governance over speed. Among U.S. respondents, that governance-first figure drops to 4%. Read the full survey findings for the breakdown.
What this means for your business
The 4% figure is the one worth sitting with. If you’re a CFO in the U.S. and you’re taking governance seriously before shipping agents into tax, compliance, or financial close workflows, you are statistically unusual. That’s not a compliment to the other 96%; it’s a description of where competitive pressure has landed the profession. The 30% of respondents whose organizations haven’t updated internal controls in the past year aren’t laggards on adoption, they’re laggards on accountability, and those are different problems with different remedies.
The survey’s framing, coming from a vendor that sells agentic compliance software, naturally tilts toward dramatizing the governance gap rather than, say, the possibility that lightweight controls are sometimes adequate. But the specific numbers are hard to dismiss on those grounds. When 44% of finance leaders say they’re only “somewhat confident” they could explain an agent’s actions to an auditor, and 23% say accountability for a significant AI error would belong to no one, the problem isn’t a vendor exaggerating risk to sell guardrails. The problem is that agents are making or recommending consequential decisions inside regulatory regimes that were built around human sign-off, and nobody has updated the org chart to reflect that.
The ROI picture complicates the speed narrative further. Only 38% say they’re generating returns at scale; 50% report limited ROI; 12% see none yet. So finance functions are absorbing governance risk to chase returns that most of them haven’t materialized. The CFO who keeps defending a slower, more deliberate rollout isn’t behind, she’s the one whose audit posture will survive the first significant agent error in a peer’s organization. I’d revise that view if a meaningful cohort of fast movers could show auditor-ready explainability alongside their scale numbers, but the survey suggests that cohort doesn’t exist yet.
Concept deep-dive: Agentic AI
Agentic AI refers to AI systems that pursue multi-step goals autonomously, deciding what to do next based on intermediate results rather than responding to a single prompt. Think of it as the difference between asking a calculator for a number and hiring an analyst who reads the filing, flags the discrepancy, drafts the memo, and routes it for approval without being asked. In finance, that autonomy is the value proposition and the governance problem at exactly the same time.
Based on reporting from CFOs struggle to get a handle on agentic AI, originally published 2026-07-22 08:42:00.
