Latest
AI Affairs, home

Wednesday 18 November 2026

Corporate

KPMG puts more than 100 AI agents to work on a parallel FTSE 250 audit

The firm plans to test the approach on up to 20 larger audits, but did not use the agents’ work to sign the pilot company’s accounts.

KPMG Auckland offices with blue glass facades beside a broad street
Photo: Prosperosity, CC BY 4.0, via Wikimedia Commons (cropped)

KPMG put more than 100 AI agents through an audit of an unnamed FTSE 250 company alongside its accountants and plans to extend the approach to up to 20 larger audits, Streamline reported on 18 November. The UK pilot, called Project Exponential, covered every phase of the audit. An audit partner approved the experimental version, but KPMG signed the company’s accounts without relying on it.

Key points

  • The agents gathered and analysed data, recorded their work and drafted audit conclusions.
  • KPMG’s UK audit chief said they sometimes questioned the company’s finances more deeply than the conventional team, but a person had to correct one failed check.
  • The UK audit regulator issued guidance on 16 November about independence risks when auditors share technology-derived insights with clients.
  • PwC has also run parallel audits to assess AI across the full process without using its work for the audit opinion.

Project Exponential covers the full audit

A team built the agents in four months. They collected information, analysed and documented it, then prepared observations and draft conclusions. The work went beyond using AI for individual audit tasks, which firms had already done.

Cath Burnet, KPMG’s UK head of audit, said the agents sometimes pursued questions about the client’s finances more deeply than the accountants did, SBS News reported. In one part of the pilot, however, a person intervened because an agent could not match regulations against company documents.

The agents’ development and operation involved two partners, one employee at directorship level, and both senior and early-career personnel. Employees assigned their objectives, supplied information about the company and earlier audits, reviewed the work periodically and challenged the conclusions. Burnet said the project team differed in size and composition from a conventional audit team.

KPMG says its system records how agents interact with documents, including the characters they use and the time of each interaction. The firm sees opportunities to save time, but declined to say whether AI-based audits would carry lower fees, Traders Union reported.

FRC sets out independence risks

The Financial Reporting Council, the UK audit regulator, published guidance on 16 November about maintaining auditor independence when using technology. It warned that sharing insights with a client’s management could threaten integrity, objectivity or independence, particularly if the work behind those insights falls outside the audit, Streamline reported.

The guidance also addresses giving clients access to auditors’ technology. The FRC prepared it in response to questions from auditors about applying its Ethical Standard. Mark Babington, the regulator’s executive director of regulatory standards, said auditors remain responsible for their use of the tools and must be able to explain and challenge their outputs. KPMG gave the FRC a high-level account of Project Exponential.

KPMG targets up to 20 larger audits

PwC has conducted several parallel audits to assess how AI performs across the entire process without relying on it, Andy Hammond, its global assurance leader, said. KPMG intends to take its model to up to 20 larger audits and eventually offer clients AI-based audits. KPMG International has contended that its own auditor should charge less because AI makes the job cheaper.

Audit committees have expressed reservations. In the Centre for Public Interest Audit’s 2026 Audit Trust Index, published in October, 23 per cent of the 57 audit committee chairs surveyed reported limited confidence in how AI was used in their company’s audit. The policy and research group’s survey covered 257 senior finance professionals between 3 and 25 August, Streamline reported.

Across the survey, 46 per cent cited over-reliance on AI weakening professional scepticism as a risk. For reviewing large volumes of data or contracts, 67 per cent expressed more confidence in AI, against 14 per cent for auditors. Respondents placed more confidence in auditors for judgement tasks.

Burnet said the technology could not yet support thousands of audits. She also said few companies had data clean enough for agents to process effectively, and that some clients might prefer not to use the service.

Sources

Topics: Agents, Enterprise adoption, Workforce

What should we cover? Tell us

The weekly briefing

The AI Affairs weekly briefing: the week's 10 most important AI stories in business, government and markets, every Monday. We'll email you a link to confirm. Every email has a one-click unsubscribe. Privacy policy