Finance Leaders are Still Wary of Too Much AI

Finance leaders are still wary of too much AI, and the numbers back it up. A fresh benchmark from spend management provider PEX, drawn from 687 finance and operations leaders, shows a workforce that is curious about AI but nowhere near ready to hand it the keys. Interest in AI in finance is close to universal, yet only a minority of teams have moved past pilots, and fewer still are comfortable letting a system make a call without a person checking the work.

That gap says less about the technology itself and more about what finance leaders need before they trust it:

• Accuracy they can verify;

• Oversight they can exercise;

• And a roadmap that doesn't ask them to leap before they're ready.

Why Finance Leaders Are Still Wary of AI in Finance

AI adoption in finance has moved quickly on paper. 31% of respondents in the PEX survey say they already use AI somewhere in their finance function, and interest in individual capabilities runs as high as two-thirds of the field (66%). But usage that broad is deployed across the whole finance operation for less than one in ten organizations. Most finance leaders are experimenting in a corner of the department, not running AI agents across the books.

The hesitation isn't a rejection of finance AI as a category. It's a rational response to what's actually at stake: routine finance tasks like coding transactions or flagging anomalies are one thing to hand off. But approving payments, closing the books, or generating financial reports without a human in the loop is another matter entirely. Finance leaders and AI don't yet share the same threshold for what counts as "good enough."

Trust in AI Accuracy Remains the Core Barrier

Ask finance executives why AI adoption is stalled and the answer is consistent across every company size and industry: trust in AI accuracy. Respondents named it the single biggest obstacle to expanding AI at 39%, well ahead of cost, integration headaches, or staff resistance. Only 28% of finance leaders said they'd be comfortable letting AI decide routine financial matters on its own, while nearly four in ten said the opposite.

That discomfort doesn't fade with scale, either. Even the largest organizations in the survey, the ones with the most mature AI programs and the biggest budgets, reported some of the highest discomfort levels with autonomous AI decision-making at 48%. Bigger deployments raise the stakes of a single wrong output, so as automation spreads, the appetite for human oversight tends to grow right alongside it rather than shrink.

AI Decision-Making Needs a Human Checkpoint

The pattern that emerges from the data is less "AI versus human" and more "AI plus human review." Finance leaders are comfortable letting AI draft, sort, and flag, but they still want a person to sign off before money moves or a number goes on a report. That's the essence of human-in-the-loop AI, and it's likely to remain the default operating model until AI reliability and transparency close the gap that trust in AI accuracy currently leaves open.

The Want-Versus-Have Gap in AI Adoption

One of the more telling findings in the PEX data is how far demand for AI outpaces actual usage across specific capabilities. Cash-flow forecasting shows the widest gap of any tool tested: about 65% of finance leaders want it, but only roughly one in ten currently use it. Audit documentation automation and fraud and anomaly detection show similarly wide gaps between interest and reality, and even AI-generated financial reports, a capability nearly half of teams want, is in active use at fewer than one in six organizations.

This want-versus-have gap is worth sitting with, because it undercuts the idea that finance teams are simply AI-averse. They're not. They want cash-flow forecasting, they want automated audit documentation, and they want AI-generated financial reports. What's missing is the AI readiness, proof, and governance that make teams comfortable turning interest into deployment.

Company Size and Industry Shape AI Readiness

The PEX benchmark also breaks its findings out by company revenue and industry cluster, and the pattern complicates the usual narrative that bigger finance teams are simply further along. Adoption climbs steadily with company size, from roughly one in five organizations under $10 million in revenue up to about half of those above $250 million. But comfort with AI decision-making doesn't climb the same way. The largest organizations in the study are also among the most cautious about autonomous AI, suggesting that as finance operations scale in complexity, governance concerns scale right alongside them.

Industry tells a similar story:

• Professional and managerial services firms lead every execution measure, running nearly twice the industry average number of live AI capabilities per team.

• Nonprofits and media and entertainment organizations report some of the lowest adoption and ROI figures despite showing just as much interest as everyone else.

This gap points to budget and staffing constraints rather than skepticism.

Mid-Market Finance Teams Are the Tipping Point

Perhaps the most interesting finding for CFOs watching enterprise AI trends is what happens in the $50 million to $250 million revenue band. This is where organizations report the highest share of teams past the experimentation stage and the strongest point-of-transaction policy enforcement of any revenue band. It's the segment where AI implementation challenges start giving way to operational discipline, and it's likely to be the band that pulls the broader market average upward as it matures.

Barriers to AI Adoption Beyond Trust

Trust in AI accuracy tops the list of AI barriers, but it isn't the only one. Interoperability and system integration issues rank second, cited by one in five (20%) finance leaders. That's a familiar problem for any team working across a patchwork of accounting software, banking platforms, and spreadsheets: connected finance and cloud finance tools only deliver value when they actually talk to one another. Cost of implementation, audit and compliance concerns, and internal resistance trail behind, which suggests most finance leaders aren't holding out for a bigger budget. They're holding out for a system they can verify and one that fits into what they already run.

Building AI Adoption One Step at a Time

PEX organizes its recommendations into a three-stage AI maturity model, often described as "Crawl, Walk, Run," emphasizing gradual adoption over waiting for a perfect AI strategy.

1. Teams at the earliest "crawl" stage, still nearly half of all respondents, are advised to automate a single, low-lift task, like receipt capture or transaction categorization, rather than attempting a full rollout on day one; 46% of crawlers haven't deployed anything yet, which makes that first capability the place every AI implementation actually begins.

2. Teams in the "walk" stage, roughly 45% of the field, are encouraged to stack a second and third capability, such as fraud and anomaly detection or real-time spend controls, while pushing policy enforcement upstream to the point of transaction instead of catching problems after the fact.

3. Only about 9% of finance teams have reached "run," where forecasting and predictive controls come into play alongside the formal AI governance needed to let automation scale without losing internal controls.

The Payoff for Finance Teams That Take the Leap

The data makes a strong argument for moving past hesitation once the basics are in place. Teams that have reached the most advanced stage of AI maturity are roughly 19 times more likely to report measurable ROI than teams that haven't started, and they report cutting manual review workloads and closing the books faster at far higher rates than teams still in the early stages. Comfort with AI decision-making rises in step with maturity too, climbing from under a fifth of "crawl" stage teams to about half of "run" stage teams, while the share still naming trust as their top barrier drops significantly.

That correlation points to something useful for any finance leader building an AI roadmap: AI reliability and AI governance aren't separate from execution, they're built through it. Comfort tends to follow proof, not the other way around, which is exactly why treating AI implementation as a staged, measurable process, rather than an all-or-nothing bet, tends to produce both stronger financial reporting outcomes and steadier internal controls.

Building a Responsible AI Strategy for Finance Operations

For CFOs and finance executives shaping their next AI roadmap, the PEX findings point toward a few practical priorities.

• Lead with AI transparency and explainability rather than autonomy, so every output can be traced and checked.

• Target the manual work that's already consuming the most hours, since receipts, coding, and reconciliation remain the biggest operational drag across nearly every company size and industry.

• And move policy enforcement toward real-time, point-of-transaction controls, since the vast majority of finance leaders still catch spend violations only after the money is already gone.

None of that requires finance teams to resolve every question about AI risk management before getting started. It requires picking one task, automating it, measuring the result, and using that evidence to build the case for the next step. Responsible AI in finance isn't about closing the trust gap overnight; it's about narrowing it one verified capability at a time, until the caution that defines this moment gives way to confidence grounded in results rather than promises.