Her research highlights a key point: audit quality depends not only on the technology being used, but also on how people work with it.

Who prepares AI-assisted advice?

Auditors often rely on valuation specialists when assessing complex estimates. These teams increasingly combine human expertise with AI. In ongoing research with colleagues, Bibler examines whether auditors respond differently depending on which member of the team prepares an estimate and which one reviews it.

In an experiment with 126 auditors from 3 Dutch audit firms, participants received the same memo from a human-AI specialist team. The memo recommended a more conservative valuation than the one proposed by the client. The only difference was whether the human or the AI had prepared the estimate, and which had reviewed it.

Auditors relied less on the advice when AI had prepared it and a human had reviewed it. As a result, they were more likely to favour the client’s higher valuation. They did not see the AI as less capable. Instead, they viewed the human reviewer as less competent and too dependent on the AI.

The findings suggest that auditors did not distrust the AI itself. They questioned the human’s role in checking its work. The effect was strongest among auditors who were less open to new ideas and less comfortable with uncertainty. The research also points to the importance of clearly explaining who did what in AI-assisted work.

For audit firms, this may mean setting out the roles of people and AI clearly, and documenting the human expert’s contribution. The way AI is integrated into a process, and how that process is communicated, may influence how much auditors rely on its output.

This study is part of Bibler’s ongoing research. Its working paper has not yet been peer reviewed, so the findings are preliminary.

Technology and professional judgement

The study builds on Bibler’s dissertation research, which examined how digital tools shape auditors’ work. It found that auditors encouraged to think innovatively performed better in a data-analytic environment. They detected fraud more effectively, offered clients more useful insights and worked more efficiently.

The research also explored how the device auditors use can affect their work. Advice written on a smartphone was less critical and sceptical than advice written on a laptop, particularly when auditors were closely involved in the task. Together, these findings show how technology, working conditions and human judgement can interact to affect audit quality.

Bibler is also working on projects that examine how auditors assess a company’s ethical climate, including when an AI tool’s analysis of management’s language differs from auditors’ own impressions of management. Another project, with UvA colleague Ulrike Thuerheimer,  explores how audit firms’ use of technology relates to team composition and audit quality.

Preparing accountants for an AI-enabled profession

For Bibler, the growing use of AI makes professional judgement more important, not less. Auditors need to assess AI-generated information they use in their work, as well as information from clients who may themselves use AI. AI can augment analysis and judgement, but its output can also be wrong. Auditors need critical thinking skills and subject knowledge to assess whether information is reliable, and where AI can usefully support their work. As AI takes on more routine tasks, junior auditors may need to act as reviewers early in their careers, which makes these fundamentals even more important.

That is also part of how Bibler teaches her Master's Auditing course. Students encounter AI throughout the course, from discussing research and ethical issues to using AI tools and reviewing their output. They need to learn how to use these tools responsibly, while also developing the fundamentals that allow them to critically assess what the tools produce.

Other colleagues in the UvA Accounting section are also researching AI from different perspectives. In separate studies, Bei Shi and Loris Haş (with Sander van Triest and Brendan O'Dwyer) study how the way AI is presented to employees and managed in organisations shapes how people use it. Ana Mićković takes a different approach, using AI as a method to study financial reporting fraud and sustainability communication. Her work also considers how generative AI could be used to make fraud harder to detect.

As AI becomes more common in accounting, the challenge is not simply to design effective tools and facilitate adoption. It is also to understand how they affect judgement, trust and accountability. Human expertise remains essential to evaluating AI’s output and ensuring that technology supports reliable information, as people, not tools, are ultimately responsible for the final judgement.