AI tools in Audit World – FRC’s New Guidance

The Financial Reporting Council has officially published guidance on the responsible use of AI in audits — providing principles-based, scenario-driven, and forward-looking direction for audit firms and professionals.

I really appreciate this guide. It’ll help auditors integrate AI tools into their toolbox. But as with other CAATs, it’s essential to understand the advantages and disadvantages of this new toolset. It could be flawed input data leading to misinterpretations, AI hallucinations generating misleading outputs, or even security vulnerabilities if the AI functions are deployed without proper controls in place.

What's in This Guidance

✅ Realistic audit use cases: The FRC walks through how to apply unsupervised ML and deep learning in fraud detection

✅ Two-tier documentation guidance: Internal files (architecture, training data) and audit files (why, how, outcomes)

✅ Explainability emphasis: SHAP, LIME, and other explainable AI methods suggested to clarify model decisions

✅ Support for in-house and third-party tools: Encourages independent assurance when vendor transparency is limited

✅ Alignment with UK AI Principles: Safety, transparency, fairness, accountability, contestability

Thematic Review Highlights:

✅ Tools must be integrated into firm methodology

✅ Documented controls and training are required

✅ Certification committees review each tool pre-approval

✅ Independence in assurance reviews is emphasized

🔗 Link to the guidelines:

https://lnkd.in/dgQUb3-w

https://lnkd.in/d8d88kKr

Author: Sebastina Burgemejster