KPMG finds finance AI use rises to 75% on governance

KPMG has published a global survey showing that active AI use in finance has risen to 75%, up from 30% in 2024. The study also found that assurance readiness is linked to stronger AI performance.

The survey drew on responses from 1,013 senior finance leaders across 20 countries and 13 sectors. It covered organisations with annual revenue of at least USD $250 million, or USD $500 million in the United States, and examined how finance teams are deploying, measuring and governing AI as the technology moves beyond trial use.

Nearly three quarters of respondents, or 71%, said AI was meeting or exceeding return on investment expectations. Leaders also reported gains in decision-making quality, speed and forecast accuracy, suggesting finance teams are using AI beyond routine automation.

Some 70% of organisations said decision-making quality had improved, while 71% reported faster decision-making. A further 64% said forecast accuracy had improved over the past year.

Sector divide

Results varied widely by industry. Banking stood out as one of the strongest performers, with 71% of leaders reporting moderate or significant improvements in forecast accuracy.

Healthcare lagged behind, with 44% reporting similar gains. KPMG linked the gap to differences in data foundations, with banks typically working from more structured data sets while healthcare organisations often deal with fragmented sources.

The figures point to a broader divide in AI outcomes across finance functions. Where data is more consistent and easier to integrate, organisations appear better placed to use AI in planning, forecasting and risk assessment.

Controls and governance

The clearest performance gap emerged between organisations that are assurance-ready and those that are not. In the survey, assurance-ready organisations were defined as those able to produce audit evidence and explain how AI-driven outcomes are reached.

These organisations reported much stronger improvement in error reduction, at 33% versus 6% for peers without the same level of readiness. They were also more confident about scaling AI, with 42% expressing confidence compared with 14% of others.

The findings suggest controls, governance and auditability are becoming more important as AI is embedded in core finance processes. The ability to show how a result was produced may matter as much as the result itself, especially when finance teams are dealing with sensitive reporting and risk decisions.

Yet fewer than half of organisations surveyed, or 42%, were fully assurance-ready for AI-enabled finance processes. Only 29% said they tracked where AI adoption fails, leaving many firms with limited visibility into where errors, bias or process weaknesses might emerge.

That gap could become more important as external scrutiny increases. Regulatory and audit expectations are pushing finance teams to show that AI tools are controlled and explainable, rather than simply effective in isolated use cases.

Data quality was another central theme in the survey. More than a third of respondents, or 36%, said it was both the biggest barrier to AI success and the biggest opportunity to improve outcomes.

This places data management at the centre of AI performance in finance. Problems with integration and system interoperability can limit the usefulness of AI models, while better data quality can expand the range of finance tasks where AI can be trusted.

Skills response

Most organisations said they were trying to address these issues by training existing teams. The survey found that 38% were focused on upskilling finance and internal audit staff, while 28% were hiring for new skills.

Organisations making the strongest progress were combining both approaches, building internal understanding of AI tools while also bringing in people who can assess data quality and interpret model outputs.

Sebastian Stöckle, Global Head of Audit Innovation & AI at KPMG International, said the issue was not simply whether finance teams adopt AI, but whether they can prove its results are reliable.

"AI only delivers real value to finance functions when it can be explained and controlled. Assurance readiness is the difference between sustained performance and hidden risk, because it turns AI into a reliable driver of results. As regulatory expectations rise, organizations that aren't ready can risk slowing down innovation instead of accelerating it. Those who invest in controls and governance early during innovation will be far better positioned to maintain trust with stakeholders," Stöckle said.

The survey covered respondents in finance, risk, audit, technology and general management roles. Technology and financial services accounted for 58% of the sample, making those sectors a large part of the overall picture presented in the findings.

The results reflect a shift in finance departments from experimenting with AI to wider operational use, with data quality and assurance emerging as the factors most likely to determine whether those deployments hold up under scrutiny.