US audit firms shift focus from AI adoption to oversight

Research from IDC, sponsored by Caseware, found artificial intelligence is now embedded across US audit and accounting firms, with the industry's focus shifting from adoption to oversight and control.

The survey found 66% of US firms have embedded AI into firm strategy, are using it in selected functions, or have pilot projects under way. A further 53% plan to adopt or expand their use of AI technologies within the next two years, suggesting continued investment after initial roll-outs.

The findings point to a shift in how the profession views the technology. Rather than treating AI as a separate project, many firms now use it in routine processes such as data analysis, documentation, and large-scale transaction review.

That change appears to be affecting how firms measure progress. The question is no longer simply whether a firm has adopted AI, but whether its use fits established workflows and professional standards.

Quality and judgment

Attitudes towards AI remain mixed even as adoption spreads. Across the global study, 53% of respondents agreed or strongly agreed that AI can enhance audit quality, rising to 63% among US respondents.

At the same time, concerns about professional judgment remain. Globally, 36% of respondents said AI could risk undermining professional judgment, and the same share was recorded in the US sample.

Those results suggest firms see benefits in broader coverage and the ability to identify anomalies, while still recognising limits on the use of automated outputs in professional work. In audit and accounting, that balance matters because responsibility for conclusions remains with the practitioner rather than the software.

Validation focus

The strongest consensus in the findings concerned checking AI-generated output. Globally, 64% of respondents said auditors should always validate AI outputs relied upon in reaching professional conclusions.

In the United States, that figure was higher at 73%, making US respondents the strongest supporters of validation among the markets covered by the study. The survey included decision-makers in Australia, Canada, Germany, the Netherlands, the UK and Ireland, and the United States.

The figures suggest that as AI becomes more common, governance frameworks and internal controls are becoming more important. Firms appear to be paying closer attention to system reviews, output testing, and how the technology fits professional obligations.

Skills shift

The research also examined which skills will matter most as AI becomes more widely used in the profession. In the US, 38% of respondents identified data analysis and interpretation as the most critical skill for accounting professionals in an AI-driven environment.

That was five percentage points above the global result of 33%. Technology and AI literacy was selected by 28% of US respondents, matching the global figure.

The combination suggests firms expect technical understanding and analytical judgment to complement each other rather than compete. As AI tools become more common in audit and accounting work, practitioners may spend less time on repetitive processing and more time interpreting results and assessing their reliability.

For Caseware, which sponsored the research, the findings reflect a market in which AI has moved into routine use. The company provides software for assurance and financial reporting and said the profession is now grappling with how the technology should be governed.

"In the US, AI is no longer something firms are piloting on the side. It is being built into how work gets done," said David Marquis, chief executive officer of Caseware. "This is why Caseware has built the first domain-specific AI platform for assurance and financial reporting - to enable firms' future growth."

The survey was conducted in December 2025 and drew responses from more than 1,000 audit and accounting decision-makers across several markets. Its results suggest the debate in the US profession has shifted from whether AI should be introduced to how its use should be supervised and validated.