# KPMG Ran 100 AI Agents Through a Full FTSE 250 Audit but Didn't Rely on Them, FT Says

Mohit Tyagi · 2026-11-18 · Enterprise

[https://timesof.ai/2026/11/kpmg-project-exponential-100-ai-agents-parallel-ftse-250-audit](https://timesof.ai/2026/11/kpmg-project-exponential-100-ai-agents-parallel-ftse-250-audit)

> KPMG deployed more than 100 AI agents this year to run a full parallel audit of an unnamed FTSE 250 company, alongside its conventional audit, the Financial Times reported on November 18. UK head of audit Cath Burnet said the agents sometimes asked better questions and dug deeper than the human team, but a person had to intervene when one agent failed to match regulations to the company's documents. An audit partner signed off the experimental audit, but KPMG did not rely on it for the company's accounts. KPMG plans to extend the pilot, called Project Exponential, to up to 20 larger audits, as the UK regulator warns about AI and auditor independence.

Audit is a business built on billing junior hours, which is why a Big Four firm testing agents on the whole job matters more than another productivity pilot. **KPMG** deployed more than **100 AI agents** this year to audit the financial statements of an unnamed **FTSE 250** company, running the experiment in parallel with a conventional audit by its accountants, the Financial Times reported on **November 18**. The firm says the agents sometimes did better than its people. It also did not trust their work enough to use it.

## What KPMG tested

In the experiment, code-named **Project Exponential**, agents working on different parts of the audit operated autonomously with one another and suggested next steps, according to the FT. They gathered, analysed and documented data, then formed draft conclusions and observations. A team built the agents in four months, and an audit partner then "signed off" the experimental audit, which KPMG did not rely on when it formally signed off on the company's accounts. Auditors had already used AI for specific tasks, but it had not been used across every phase of a large UK audit, the FT reported.

KPMG would not disclose the size of the team, saying only that two partners and one director worked with senior managers and junior staff to build and run the agents. Staff set the agents' objectives, supplied information about the business and previous audits, checked their work periodically and challenged their conclusions.

## Claims and checks

The performance claims are KPMG's own. "In several cases the AI was able to go so much deeper" than the regular audit team, UK head of audit **Cath Burnet** told the FT, adding that in some situations it had asked "better questions". In one area, though, a human had to intervene after an agent failed to match a set of regulations to the company's documents. KPMG acknowledged "clear time-saving opportunities" but declined to say whether fees would fall. Burnet said its system records the agents' interactions with documents down to individual characters and timestamps.

No outside party has assessed the results. KPMG gave the Financial Reporting Council a high-level overview of the experiment, the FT reported. The FRC warned this week that using AI could pose a "distinct threat" to auditors' integrity and independence if they used insights from AI agents to advise clients; "Whatever technology a firm uses, the auditor remains accountable," said Mark Babington, its executive director of regulatory standards. KPMG has been here before with AI-generated material: in June it [pulled an agentic AI report](https://timesof.ai/2026/06/kpmg-pulls-agentic-ai-report-alleged-hallucinations) after companies named in it disputed its case studies.

## What comes next

KPMG plans to expand the project to up to **20 larger audits** and hopes eventually to offer AI-based audits to clients. Burnet said the technology is not yet available at the scale needed to run thousands of audits, that some companies may prefer not to use the service and that few have data "clean" enough for agents to work with. Rival **PwC** has run several parallel audits of its own to test AI without relying on it, its global assurance leader Andy Hammond told the FT.

## Limits

- Every detail comes from the FT's interview with KPMG; the audited company is unnamed and KPMG has published no results.
- "Better questions" and "deeper" are KPMG's characterisations, with no measure of accuracy or error rates disclosed.
- The parallel audit was not used for the real opinion, so it shows what agents produced alongside humans, not what they would produce alone.

## Sources

- [Financial Times via The Irish Times: AI asked 'better questions' than KPMG accountants in pilot audit of effectiveness (November 18, 2026)](https://www.irishtimes.com/business/2026/11/18/ai-asked-better-questions-than-kpmg-accountants-in-pilot-audit-of-effectiveness/)
- [Business Standard: KPMG finds AI agents ask better questions than human auditors in trial (November 18, 2026)](https://www.business-standard.com/industry/news/kpmg-finds-ai-agents-ask-better-questions-than-human-auditors-in-trial-126111800471_1.html)
