Skip to content
AI EmployeeSuper-AgentsAgent-to-AgentBenchmarksPricingBlog

KPMG's First AI Audit: 100 Agents on a FTSE 250 Company

At a glanceQuick answers
What did KPMG do?
It ran more than 100 AI agents through a full parallel audit of an unnamed FTSE 250 company, an internal project it calls Project Exponential.
Did the AI sign off the accounts?
No. The official audit opinion was based on the regular audit, run by people.
What comes next?
KPMG plans to expand the project to as many as 20 larger audits, according to the FT’s report.
Data illustration on a near-white ground: two parallel lanes run to a stack of company accounts, a thin lane of human figures and a dense lane of a hundred small teal blocks, with a huge black 100 and one orange pen on the signed report.
Fig 0Two audits of one company, one signature. Made by CellCog's image agent, running GPT Image 2.6.

KPMG has run more than 100 AI agents through a full audit of a UK-listed company for the first time. The Financial Times reported it on November 18, 2026, and its post on X carried the line that “the AI was able to go so much deeper” than the regular audit team.

The audit was a parallel one: the agents worked through every phase of the annual audit of an unnamed FTSE 250 company while the regular team did the real audit, and the signed opinion came from the regular audit. We read the FT’s post on X, the-decoder’s report, which cites the FT at length, and the regulator’s own release. We did not open the FT article.

On this page · 6 sectionsOpen
  1. What KPMG ran
  2. What the regulator said two days earlier
  3. How the story spread
  4. What is not established
  5. Where CellCog sits
  6. Sources
Key points5 · 5 min full read
  1. Two parallel arrows running to one document.
    KPMG ran more than 100 AI agents through every phase of an experimental audit of a FTSE 250 company, in parallel with the regular audit, the FT reported on November 18, 2026.
  2. A magnifying glass over a ledger.
    KPMG’s UK head of audit, Cath Burnet, said that in several cases the AI went deeper than the regular team and asked better questions.
  3. A hand stopping a small block.
    A human had to step in when an agent tried to match a set of regulations against the company’s documents.
  4. A pen over a signed report.
    The signed audit opinion came from the regular audit. KPMG plans to expand the project to as many as 20 larger audits.
  5. A balance scale.
    Two days earlier, the UK regulator, the FRC, published guidance on auditor independence when using AI.

§ 01What KPMG ran

What As reported
Company An unnamed FTSE 250 company
Agents More than 100, across every phase of the audit
What the agents did Collected, analyzed and documented data; drafted audit conclusions
Where a human stepped in Matching a set of regulations against company documents
Signed opinion Based on the regular audit
Next step As many as 20 larger audits
Table 1KPMG’s Project Exponential, as reported by the FT and the-decoder, November 18, 2026

Cath Burnet, KPMG’s UK head of audit, said that in several cases “AI was able to go so much deeper” than the regular team, and that in some situations the technology asked “better questions” than KPMG’s own auditors. KPMG’s internal name for the work is Project Exponential.

The people stayed in the loop. KPMG staff set the agents’ goals, gave them context, reviewed their work and challenged their conclusions before the opinion was final, and the-decoder reports that “A human had to step in when an AI agent tried to match regulations against company documents.” KPMG says the system gives “full visibility and full traceability”, recording each agent’s work on the documents down to single characters and timestamps.

Burnet also set a limit: the technology is not ready for thousands of audits, because few companies have data clean enough for agents to work with reliably.

Animation: a grid of small agent blocks lights teal in a sweep toward a stack of company accounts, whose top ledger turns teal, while three pillars for the regular audit team stand still and an orange tick is drawn on the signed opinion card at the end of their lane
Fig 1A crowd of AI agents and a small audit team reach the same accounts; people sign

§ 02What the regulator said two days earlier

The Financial Reporting Council published guidance on November 16, 2026 on auditor independence when using technology, including AI. It does not name KPMG. Its concern is what an auditor tells a client: AI agents that read far more of a company’s data than a sample can surface findings well beyond the audit, and the-decoder reports that the FRC sees sharing some of them as a “distinct threat” to independence.

Mark Babington, the FRC’s executive director of regulatory standards, said in the release that “auditor independence is and must remain at the heart of audits providing value and serving the public interest.” The-decoder quotes him more bluntly: “Whatever technology a firm uses, the auditor remains accountable.”

§ 03How the story spread

Time (UTC) Account What it was Views
Nov 18, 5:31 @FT The exclusive 594,261
Nov 19, 4:47 @taxologyin A summary for tax professionals 2,603
Nov 19, 6:00 @MwangoCapital A market summary 2,426
Nov 19, 6:27 @aptvibes__ A summary post 2,604
Table 2Posts about KPMG’s AI audit on X, times computed from post ids, views read through the X API between 08:43 and 08:45 UTC, November 19, 2026
Views on X, read 08:43 to 08:45 UTC, November 19, 2026Bar chart of views: the FT's exclusive highlighted, far above three summary postsFT exclusive594261aptvibes summary2604taxologyin summary2603MwangoCapital2426Views on X, read 08:43 to 08:45 UTC, November 19, 2026Bar chart of views: the FT's exclusive highlighted, far above three summary postsFT exclusive594261aptvibes summary2604taxologyin summary2603MwangoCapital2426
Fig 2Views on X, read 08:43 to 08:45 UTC, November 19, 2026

The FT’s post held almost all of the attention; most other posts repeated its two quotes.

§ 04What is not established

  • Which company was audited, and what the agents found that the regular team missed.
  • What audit fees do next. KPMG sees “clear time-saving opportunities”, in the FT’s account, and would not say whether fees will fall.
  • How the other Big Four compare. The-decoder reports that PwC has already run several parallel audit projects of its own, with no results published.

§ 05Where CellCog sits

CellCog builds AI employees, so the shape of this result is familiar: agents doing the volume, people setting the goals and signing. We do not sell audit software, and nothing here is advice on audits. ChatGPT launched a plugin for accounting work the same day, and our ranking of AI bookkeeping software covers the tools for the books before they reach an auditor.

§ 06Sources

Frequently asked5 questions

Q1What is KPMG's Project Exponential?

KPMG’s internal name for an experimental audit in which more than 100 AI agents worked through every phase of a FTSE 250 company’s annual audit, alongside the regular audit team.

Q2Did the AI agents do better than human auditors?

In parts. Cath Burnet, KPMG’s UK head of audit, said that in several cases the AI went deeper and asked better questions. In one area a human had to step in when an agent could not match regulations against the company’s documents.

Q3Which company was audited?

KPMG has not named it. The reports describe it only as a company in the FTSE 250.

Q4Will AI audits lower audit fees?

KPMG sees clear time-saving opportunities, according to the FT, but would not say whether fees will fall.

Q5What did the FRC say?

On November 16, 2026 the FRC published guidance on auditor independence when using technology, including AI. Its executive director of regulatory standards, Mark Babington, said the auditor remains accountable whatever technology a firm uses.

Published 19 November 2026 All Workflows & use cases →