US CFOs back AI for finance but insist on oversight

A survey of mid-market US Chief Financial Officers has found broad interest in adopting AI for finance work, but limited trust in AI outputs without human checks.

Wakefield Research interviewed 100 CFOs at US companies with annual revenue between USD $50 million and USD $500 million. The results pointed to a gap between perceived value and willingness to rely on AI for accounting data.

According to the study, 96% of CFOs said AI's biggest benefit lies in freeing up time for strategic work. Only 14% said they completely trust AI to deliver accurate accounting data on its own. The survey also found that 97% of CFOs said human oversight remains critical.

The findings reflect a shift in how finance leaders frame AI adoption. The question now centres on how much automation firms can accept while still meeting audit and governance expectations.

Audit expectations

The study described finance teams as caught between two approaches. It said AI copilots still require accountants to check transactions one by one. It said this model yields limited productivity improvements.

It also pointed to AI agents that claim to automate finance processes end to end. It said CFOs associate that approach with higher risk. It cited a lack of visibility into how outputs are produced. It also cited limited audit trails and weak understanding of business context.

The research said finance leaders want systems that handle routine work and then escalate exceptions to humans with context. The study referred to this as "intelligent escalation".

One CFO described the desired end state as an autopilot model. The CFO said: "We need an autopilot - fast, accurate and with the sound judgment of our most reliable accountant."

Human in loop

The survey results suggest finance teams expect continued involvement in decision-making, even where AI handles routine processing. That stance fits with the reported 97% figure on the need for oversight.

It also signals pressure on AI vendors to offer verifiable records for automated actions. In finance operations, internal controls often rely on evidence trails. Those trails cover who made a decision, what data supported it, and what rules applied.

The study argued that accuracy and speed now sit alongside auditability as core requirements. It also framed "judgment" as a differentiator. That term covered when a system should proceed automatically and when it should pause and pass an item to a person for review.

Vendor positioning

The research referenced Maximor, which positions itself as a provider of finance automation software that works alongside existing ERP systems. The company described its approach as an "Audit-Ready Agent" architecture. It said this produces "fully verifiable outputs and decision traces for every action".

Dominic Rand, CFO, Kiva Brands, described his experience of evaluating AI products for finance teams.

"We evaluated several AI solutions, and the difference was night and day," said Dominic Rand, CFO, Kiva Brands. "Most tools either wanted full control with zero transparency, or they created more work for my team. What we needed was what Maximor delivered: AI that could handle the routine with speed and precision and knew exactly when to bring a human into the loop. That's when automation becomes a partnership, not a risk."

Ramnandan Krishnamurthy, Co-Founder and CEO, Maximor, linked the survey findings to broader trends in the AI market.

"When intelligence becomes commoditized, judgment becomes the competitive advantage," said Ramnandan Krishnamurthy, Co-Founder and CEO, Maximor. "CFOs aren't asking for smarter AI - they're asking for AI that knows its limits. They need systems that are verifiable, operate autonomously when appropriate, and demonstrate judgment about when to act and when to escalate. That's the shift we're seeing in the market."

Adoption signals

The research said between 60% and 77% of CFOs plan to adopt AI depending on the use case. It did not break out the categories in the summary, but the range suggests CFOs differentiate between tasks with clear rules and those with higher ambiguity.

Mid-market finance functions often face constraints on headcount and specialist capacity. Automation products tend to focus on accounts payable, expense processing, reconciliations, close processes, and reporting workflows. These areas also sit inside established control frameworks.

For vendors, the survey indicates that product claims about automation will face scrutiny from finance leaders who must sign off numbers and defend processes to auditors and boards. The survey's emphasis on oversight also implies that adoption could depend on how tools handle exceptions and documentation, not only on raw throughput.

The study concluded that AI systems without verifiable workings and escalation processes would not meet finance leaders' expectations as AI adoption moves from pilots into live accounting operations.