357
Productivity & Workflow354
Software Development248
Automation & Workflow238
Marketing & Growth204
AI Infrastructure & MLOps197
Photography & Imaging166
Writing & Content Creation193
Design & Creative170
Data & Analytics165
Customer Support156
Education & Learning144
Voice & Speech130
Sales & Outreach132
Operations & Admin119
KPMG ran a pilot audit using more than 100 AI agents alongside its normal audit team. It found time savings, but humans still had to step in.
In short: KPMG says AI agents asked better questions than its accountants in a pilot audit, but the firm still needed human checks and intervention.
KPMG, one of the “Big Four” accounting firms, carried out what it described as its first AI-based audit of a UK-listed company. The client was an unnamed company in the FTSE 250, which is an index of mid sized companies listed in London.
KPMG used more than 100 AI “agents” (software helpers that can handle tasks on their own) to review the company’s financial statements. The firm ran this pilot audit in parallel with a normal audit by its accountants, and it did not rely on the AI work when it officially signed off the company’s accounts.
Cath Burnet, KPMG’s UK head of audit, said the AI went deeper on some issues and in some situations asked “better questions” than the audit team. However, she said a human had to step in when an AI agent failed to match a set of regulations to the company’s documents.
KPMG said the project, codenamed Project Exponential, showed “clear time-saving opportunities,” but it would not say whether audit fees would fall. KPMG plans to expand the approach to up to 20 larger audits, but Burnet said the technology is not ready to run thousands of audits yet, and many companies do not have “clean” enough data for AI to work with.
Audits are meant to reassure investors and the public that a company’s numbers can be trusted. If AI helps auditors check accounts faster and spot more issues, audits could become quicker and possibly cheaper. But regulators warn that auditors are still responsible for the final answer, and they must be able to explain what the AI did, challenge it, and avoid relying on false outputs (like a calculator that sometimes makes up numbers).
Source: Financial Times