Overview
- KPMG ran an experimental deployment of more than 100 AI agents that collected data, analysed records, drafted conclusions and coordinated tasks across an audit workflow.
- KPMG said the agents in several cases probed issues more deeply and flagged follow-up questions that the human audit team did not raise.
- The trial was non-statutory and an audit partner signed off on the exercise but its findings were not used to approve the company’s accounts.
- Human oversight remained essential when an auditor intervened after an agent failed to match regulatory requirements to company documents.
- KPMG plans limited expansion to around 20 larger audits but said poor client data quality and the system's immaturity block mass deployment while regulators press for safeguards to protect independence and scepticism.