UK FRC’s first guidance on generative and agentic AI in audit

Global reference

• Country: United Kingdom, with wider influence on audit practice
• Regulator: Financial Reporting Council (FRC)
• Topic: Use of generative and agentic AI tools in audit engagements
• Key date: Published 30 March 2026
• Documents: Generative and Agentic AI Guidance, and a companion factsheet

What the guidance says

• Audit firms should get appropriate confidence in the quality of AI outputs before relying on them. The guidance gives a framework for this.
• How much checking is needed depends on the tool and its use, and is a matter of professional judgement.
• AI use should sit within the firm’s quality management system under ISQM (UK) 1.
• Accountability does not change. In the FRC’s words, the human auditor is always accountable.
• It is not a response to quality failures. It sets out good practice and a base for future FRC work.

Examples the FRC uses

• Summarising board minutes.
• Reviewing contracts with an AI tool for revenue recognition testing.

Why it matters outside the UK

• Audit firms are global networks, so FRC guidance shapes methods used in other countries.
• Agentic AI, meaning tools that plan and carry out steps, is now treated as an audit tool to control, not a future idea.
• Audit regulators in other countries are likely to follow.

What finance teams should expect

• Auditors may ask for data in formats their AI tools can read, such as full ledgers and contract sets.
• Expect more questions on anomalies that AI finds across full populations, not samples.
• If your team uses AI to prepare journals, reconciliations or disclosures, expect auditors to ask how you control it.

What to do now

• Agree with your auditor early which AI tools they will use and what data they need.
• Document your own AI use in financial reporting: the tool, its purpose, the reviewer and the evidence kept.
• Keep human sign-off on key judgements.

Our take

The guidance is practical and balanced. It does not slow AI in audit; it asks firms to prove they can trust it. Finance teams should apply the same test to their own AI use.