AI methods are being integrated into audit workflows

What the study found: Artificial intelligence (AI) methods are increasingly being integrated into parts of the audit workflow, including planning, risk assessment, control tests, and substantive procedures/reporting. The review also identifies a proposed AI-based audit workflow reference architecture with human-in-the-loop controls, meaning auditors remain involved in oversight and review.
What the authors say this matters: The authors conclude that the findings have practical implications for auditors, standard-setters, and system designers who are seeking to revise audit approaches and regulations to support AI-driven assurance.
What the researchers tested: The study is a systematic review of 100 peer-reviewed articles published from 2015 to 2025. The authors searched five large-scale databases and other sources, used PRISMA (a structured reporting approach for systematic reviews), extracted and assessed data, and then conducted narrative and thematic analysis.
What worked and what didn't: The review reports improvements in detection capabilities, coverage, and efficiency in various empirical and design science studies. It highlights machine learning-based anomaly detection and predictive analytics, natural language processing for document analysis, and robotic process automation for automation as approaches that are becoming part of audit work; however, it also notes technical, organizational, and regulatory obstacles that still limit broader adoption.
What to keep in mind: The abstract notes gaps in longitudinal assessment, comparative evaluation of AI methods, and regulatory recommendations. Limitations are otherwise not described in the available summary.

Key points

• The review examined 100 peer-reviewed articles on AI in auditing published from 2015 to 2025.
• AI methods discussed include machine learning, natural language processing, and robotic process automation.
• The review says these methods are being used in audit planning, risk assessment, control tests, and reporting.
• Reported benefits include improved detection capabilities, coverage, and efficiency.
• The abstract notes technical, organizational, and regulatory obstacles, plus gaps in longitudinal and comparative evaluation.