Finance leaders back AI spending in accounts payable

Basware has published commissioned research showing that finance leaders plan to increase spending on artificial intelligence in accounts payable, while holding back broader support until returns are proven. The study found that only 39% say they are ready to scale AI.

The survey of 231 finance and accounts payable technology decision-makers in the UK, US, France and Germany points to a gap between adoption and operational discipline. While 67% said their teams already use AI in targeted accounts payable tasks, 68% said they need demonstrable return on investment before approving further finance technology spending.

The findings add to a broader debate over whether AI tools can deliver measurable value within core business systems rather than through stand-alone assistants. They also reflect growing scrutiny of autonomous software in production settings, particularly where financial controls, audit trails and regulatory obligations are involved.

Investment plans

Three-quarters of respondents said they expect AI investment to rise over the next 12 to 24 months. Yet expectations on timing remain restrained, with only 7% saying they expect AI investments in accounts payable to pay back in less than six months.

Another 20% expect returns within six to 12 months, while 35% see payback taking 13 to 24 months. This suggests finance departments are treating AI as a longer-term operational project rather than a rapid cost-saving measure.

Spending is increasingly being directed towards areas where outcomes can be measured against existing controls and workflow data. In accounts payable, that includes exception handling, approval recommendations and other tasks where software actions can be compared with established processes.

At the same time, respondents signalled that reliability matters more than novelty. The study found that 64% prioritise stability and compliance over raw innovation when selecting AI tools, while only 46% said they had achieved an effective balance between governance and innovation.

Control concerns

The data suggests many finance teams remain unconvinced they have the structures needed to manage AI at scale. Although 67% reported some use of AI in accounts payable, only 39% said they operate AI centres of excellence at scale.

That shortfall is significant because accounts payable sits close to the controls that govern cash movement, supplier payments and fraud prevention. Any AI system that recommends or completes actions in that environment is likely to face greater scrutiny from finance leaders than systems used for lower-risk administrative work.

Regulatory pressure is also shaping those decisions. Some 65% of respondents said major or urgent improvement is needed to adapt to new financial regulations, and 63% cited rising demand for data-backed decision-making.

Donna Wilczek, Chief Product and Technology Officer at Basware, said finance offers one of the clearest tests of whether AI produces measurable business value. "Finance is a strong place to start with AI because the value can be measured," she said. "The challenge is getting from ambition to execution in a way the business can trust. Once outcomes are proven, the remit can grow."

Audit trail

For finance leaders, the issue extends beyond whether a model can automate a task. They also need records showing what the system did, why it acted and when human intervention was required.

Wilczek said those requirements are becoming mainstream governance expectations for AI used in finance operations. "Governed AI is no longer aspirational, it's a board-level requirement," she said. "Every AI decision in accounts payable needs to be logged, traceable, and auditable from the moment it's made, not reconstructed after the fact."

Basware is using the research to frame its approach to what it calls governed autonomy, under which finance teams define how much authority AI tools have over a process. Human review remains part of workflows where judgment is needed, and broader autonomy should follow only as outcomes are verified.

The framework sets out three levels of AI authority, moving from an advisory role to a collaborative one and then to a more autonomous operating role. That reflects a wider pattern among business software suppliers, many of which are positioning AI not as a fully independent agent but as software whose actions remain bounded by internal rules and audit requirements.

The results also come amid wider concern about how autonomous systems behave in live environments. Recent scrutiny of AI agents acting with limited oversight has sharpened questions for large companies about access controls, monitoring and accountability when software is allowed to take action rather than simply generate suggestions.

Against that backdrop, the survey indicates that finance departments are not rejecting AI, but are placing stricter conditions on its use. "Success in the next phase of AP won't be achieved by the teams using the most AI, but by the teams that govern it best," Wilczek said.