Payhawk flags AI governance gap in finance leaders

Payhawk has published research showing that 45% of finance leaders who describe their organisations as AI leaders lack minimum rules for AI use. The findings are based on a global survey of 1,520 finance and business leaders.

The research focused on a subset of 405 organisations that rated their AI maturity between seven and 10 out of 10. It found that only 26% of these self-declared leaders had all five operational requirements needed to move AI from adoption to routine use in finance workflows.

The five requirements were execution measures, minimum rules for AI use, skills and tools, a committed budget, and data suitable for AI analytics. Among this group, governance rules ranked lowest, with 55% reporting having them in place, while 78% reported strong skills and tools.

The data suggests finance teams do not progress along a single AI maturity ladder. Instead, respondents fell into six operating postures that reflect different combinations of readiness and weakness.

Six groups

The largest segment, labelled scaled adopters, accounted for 26.9% of AI leaders. These organisations were strong across all five requirements.

Incremental improvers made up 17.5% of the sample. Their AI readiness was uneven across the operating model, with no single area showing clear strength.

Execution-led implementers represented 16.0% of AI leaders. These organisations showed strong execution and skills, but lacked minimum rules for AI use.

Another 14.1% were classified as agent-first, control-later, in which experimentation and early deployment took precedence over governance and formal execution readiness. Governance-forward scalers accounted for 13.8%, with stricter rules and execution but weaker data readiness, while control-first planners made up 11.6%, showing relative strength in skills, budget and data but weaker execution measures.

The research points to two recurring obstacles behind stalled AI use in finance: rules debt and data debt. In the first case, organisations introduce AI before putting governance structures in place that allow systems to be audited, explained or used safely in approval, compliance and financial control processes.

That pattern appears most clearly in the execution-led implementers and agent-first, control-later groups. Together, those two segments account for about 30% of self-described AI leaders in the survey.

Data debt, by contrast, describes organisations that have established controls and execution processes but still lack consistent, complete underlying data. In these settings, firms may be able to oversee how AI is used but may not rely on its outputs across financial operations.

Governance gap

The findings come as finance teams face growing pressure to automate tasks while maintaining controls over spending, reporting and approvals. The survey suggests many organisations view their AI efforts as advanced, even when key governance and data foundations remain incomplete.

This creates a mismatch in investment decisions, according to Payhawk. Some organisations continue to spend on AI tools when the main barrier is governance infrastructure, while others focus on policy frameworks when weak data quality is the larger problem.

The survey covered senior professionals across finance, accounting, sales, human resources and procurement. Respondents included C-suite, vice president, director, and senior individual contributor roles across eight countries, at companies with 50 to more than 1,000 employees.

The study included respondents from DACH, Spain, France, Benelux, the UK and Ireland, and the United States. It also spanned sectors including services, digital, manufacturing, healthcare, education, non-profit and business-to-consumer companies.

For Payhawk, the results indicate that AI deployment in finance is being constrained less by technical enthusiasm than by operating discipline. Organisations that appear active in AI can remain limited to narrow assistive uses when governance is weak. At the same time, more tightly controlled organisations can still struggle to scale if the underlying data is fragmented.

"Finance AI scaling feels inconsistent because organisations are progressing unevenly across the capabilities that underpin scale", said Hristo Borisov, Chief Executive Officer and Co-Founder of Payhawk. "Many organisations are investing in more AI when the real bottleneck lies elsewhere - in rules or data. Scaling AI in finance is fundamentally an orchestration challenge: coordinating rules, data, and accountability across workflows. Those that only address some readiness requirements will face inherent trade-offs and remain stuck in assistive use cases."