Finance teams adopt AI faster than controls, Medius says

Medius has published research showing finance teams are adopting AI faster than they are putting controls around its use. The survey covered 2,386 senior finance professionals in the US, UK, Sweden and France.

The findings point to a gap between stated policy and day-to-day practice. While 90% of respondents said there is always a financial or compliance threshold requiring human approval for AI decisions, 45% said their teams often act on AI-generated recommendations without human intervention.

Another governance weakness emerged around documentation. Nearly half of respondents (46%) said AI approval thresholds are understood informally rather than formally documented, raising questions about how organisations would audit or defend decisions made with AI input.

More autonomous systems are also spreading. Medius found that 38% of finance executives already have agentic AI operating in at least some finance processes, while another 50% plan to deploy it within the next 12 months.

The research also suggests responsibility for AI-related mistakes remains unsettled. When asked who would be accountable if an AI error caused a financial loss or compliance issue, respondents were split almost evenly: 28% named IT leaders, 27% the finance leader who approved it and 27% the employee who acted on the recommendation.

Ahmed Fessi, Chief Transformation & Information Officer at Medius, addressed that lack of clarity.

"Organisations are growing comfortable allowing AI to influence decisions, yet many still lack clear ownership when those decisions go wrong," said Ahmed Fessi, Chief Transformation & Information Officer at Medius.

"As autonomous finance becomes more common, explainability, governance and accountability are just as important as accuracy," Fessi said.

Fraud concerns

The study also found signs of tolerance for low-value fraud inside finance departments. Eighty-seven per cent of finance executives said they had ignored a small expense, reimbursement or claim they believed was fraudulent.

That pattern extended to respondents' own behaviour. Some 67% said they would be likely to make a minor dishonest expense claim themselves if that behaviour were common among co-workers, while 57% said they would round up an expense or mileage claim if they believed it would go unnoticed.

Nearly three-quarters, 74%, said small forms of fraud causing minor financial losses are already common in workplaces. Another 64% said they would feel justified in committing a small dishonest financial act if they felt underpaid or undervalued.

Medius described this trend as the rise of the shallowfake, referring to low-value financial rule-bending that may appear minor in isolation but can build up over time.

Concern about more sophisticated fraud remains high. Ninety-three per cent of respondents said they are concerned about AI-generated fraud over the next 12 months, while US organisations reported average yearly losses of USD $168,000 from invoice fraud and an average of one successful invoice fraud attempt each month.

Chris Wilmot, Chief Financial Officer at Medius, contrasted headline-grabbing scams with routine losses.

"Costly deepfake fraud gets all the headlines," said Chris Wilmot, Chief Financial Officer at Medius.

"But while finance professionals keep a sharp eye out for these scams, hundreds of thousands of dollars are slipping out the back door through shallowfake fraud. These seemingly minor 'micro frauds' add up to death by a thousand cuts for organisations without the controls to catch them," Wilmot said.

Late payments

The survey also highlighted a disconnect between automation spending and payment performance. Although 85% of finance teams reported some level of accounts payable automation, nearly half of invoices still require manual intervention.

Respondents estimated that 28% of invoices are paid late in a typical month. Almost all (96%) said managing late payments has contributed to stress or burnout among accounts payable teams.

Suppliers are also responding to delays. According to the research, 45% of respondents said suppliers imposed stricter upfront payment terms, 43% said suppliers ended relationships altogether, and 42% said suppliers reduced service quality or speed. Another 42% reported disputes escalating to formal legal action.

Kevin Permenter, Research Director, Financial Applications at IDC, said the issue is no longer just adoption.

"Automation was supposed to make late payments the exception, not the norm. Yet the findings suggest many organizations continue to experience payment delays despite ongoing technology investment. As supplier terms tighten and business disruptions grow, the challenge is no longer simply adopting automation, but ensuring those investments deliver measurable operational outcomes," Permenter said.

Workforce impact

The research also pointed to AI's growing role in employment decisions. More than half of finance leaders, 55%, said AI fluency has become a meaningful differentiator in recruitment, while 86% said employees' use of AI is now a factor in performance evaluations.

At the same time, respondents reported a human cost from the shift. Three-quarters (75%) said AI use has increased worker fatigue or burnout, even as organisations continue to embed the technology more deeply into finance operations.

The survey covered senior finance professionals responsible for financial operations, technology investment and strategic finance initiatives across four countries. The results show a function under pressure from fraud, late payments, unclear AI accountability, and changing staff expectations.