Field report SW-52 / J1 · sheet /operations
Operations & RestructuringFiled from the territory
KPMG completes first AI-agent audit of FTSE 250 company
KPMG ran more than 100 AI agents through a full FTSE 250 audit, with the system going deeper than accountants in places; up to 20 larger audits are next.
- Grid
- SW-52 / J1
- Surveyed
- November 19, 2026
- Reading
- 3 min read
- Correspondent
- Elena Vasquez

Field notes
- KPMG completed its first AI-driven audit of an unnamed FTSE 250 company, run alongside a conventional audit.
- Project Exponential deployed more than 100 AI agents across all audit phases; the system was built in four months.
- Audit head Cath Burnet said the AI went "so much deeper" in several areas and asked "better questions".
- KPMG did not rely on the AI audit when approving the formal accounts; a partner signed off only the experiment.
- The firm plans to extend the trial to up to 20 larger audits before offering the approach to clients.
KPMG has completed the first AI-driven audit of a UK-listed FTSE 250 company, running more than 100 autonomous agents through every phase of the engagement alongside a conventional audit, the Financial Times reported. The company was not named.
The exercise, known internally as Project Exponential, marks the first time AI agents have been deployed across a full major UK audit rather than on isolated tasks. KPMG UK's audit head, Cath Burnet, said the AI outperformed her accountants in parts of the work. "The AI was able to go so much deeper" in several financial areas, she said, and in some cases asked "better questions" than the human team.
Where the machine fell short
The system did not run unattended. At least one auditor had to intervene after an AI agent failed to match applicable regulations against the company's documents — a gap that underscores why the firm kept humans in the loop throughout.
During the trial, the agents gathered and analysed data, documented their own work, proposed next steps, and prepared draft findings and observations. KPMG built the entire system in four months. The development team had a deliberately unconventional profile: two partners, one director, senior managers and junior accountants built and operated it.
Burnet said the team had a "different shape and size" from a conventional audit team, but stressed that human involvement remained "critical". Staff set each agent's objectives, fed them company information and prior audit files, reviewed the output, and challenged the conclusions the system produced.
What does this change for the audit market?
For now, nothing formal. An audit partner signed off the experimental audit as a controlled exercise, but KPMG did not rely on it when approving the company's actual accounts. The regulated sign-off remained a wholly human, conventional process.
The economics, however, are already moving. KPMG identified "clear time-saving opportunities" in the trial, though it declined to say whether those savings would translate into lower client fees. The pressure is nonetheless visible inside the firm's own walls: KPMG International has secured reduced fees from its own auditor by arguing that AI should drive audit costs down — a signal that clients may soon make the same argument back.
The Big Four firm intends to extend the trial to up to 20 larger audits before deciding whether to offer the agent-based approach to clients commercially.
The strategic stakes
The trial puts a number on a shift the profession has debated for years. Auditors have used AI for individual tasks — sampling, document review, anomaly detection — for some time. Project Exponential is the first time a Big Four firm has strung those capabilities into an end-to-end process on a live FTSE 250 engagement, from data gathering to drafted findings.
The four-month build time and the lean team behind it also signal how quickly such systems can now be assembled. What has not changed is the accountability structure: a partner signs, regulators watch, and the AI's work remains advisory until the profession's rules say otherwise.
For now, the machines ask better questions. The answers still belong to the humans.
via internationalaccountingbulletin.com (Source)
More from Elena Vasquez
Show full bio
Staff writer covering companies, deals and strategy in corporate strategy and chief-executive leadership at CEO Frontier.
486 articles
Adjacent sheets
- KPMG launches AI-native audit platform with 500-team rollout
- IBM warns British firms cannot prove their AI is under control
- UK Audit Watchdog Warns Firms Against Using AI to Win Advisory Work
- dnl Enters US Market With GAAP Disclosure Checklist Tool
- AI Agents Will Recommend Layoffs, Arist CEO Michael Ioffe Predicts