AI newsBusiness2 min read

KPMG audit trial shows AI agents asking deeper questions

KPMG audit trial shows AI agents asking deeper questions

Photo: Arwahyn, AI-generated illustration

The Rundown

KPMG UK revealed in a trial that over 100 AI agents investigated corporate finances more deeply and asked better questions than human auditors.

The details

  • KPMG UK deployed more than 100 AI agents to examine the accounts of a FTSE 250 company in Project Exponential.
  • Cath Burnet, head of audit at KPMG UK, said the AI agents investigated issues more deeply and asked questions human teams missed.
  • A human auditor had to intervene during the trial after an AI agent failed to match regulatory requirements with documentation.
  • KPMG plans to expand the agent experiment to up to 20 larger corporate audits.

The full story

KPMG UK revealed that autonomous AI agents examining a major UK-listed firm investigated financial anomalies more deeply than traditional staff. Codenamed Project Exponential, the trial ran over 100 agents alongside an existing audit team on a FTSE 250 company. KPMG UK head of audit Cath Burnet confirmed that the system repeatedly spotted angles that human auditors did not consider.

The software moved beyond routine administrative support into handling the entire workflow. The agents gathered data, ran deep analytical checks, drafted formal audit observations, and coordinated complex steps among themselves. Developed over four months, the multi-agent setup divided analytical responsibilities dynamically across several stages of the accounting review.

Significant technical hurdles remain before the technology can operate without close human supervision. During testing, an auditor had to step in after an agent incorrectly matched statutory regulatory requirements with company records. KPMG emphasized that the test findings were not used to sign off on statutory financial accounts, and messy enterprise data remains a recurring friction point.

KPMG plans to roll the architecture out across up to 20 larger client audits. The Big Four accounting firms rely heavily on armies of junior professionals for data extraction and compliance checks. If automated multi-agent networks successfully shoulder analytical heavy lifting, professional services firms may restructure billing models and redirect experienced partners purely toward high-level judgement.

Why it matters

The findings show that autonomous agent swarms are moving past basic back-office chores into high-stakes corporate analysis. KPMG demonstrates that coordinated AI models can sometimes outperform human professionals at scrutinizing financial records. This shift challenges the traditional Big Four business model of using large teams of junior auditors for manual verification.

What it means for you

Bangladeshi accountants and finance freelancers should expect routine data review tasks to shift rapidly toward autonomous agent platforms. Learn how to prepare clean data pipelines and craft multi-agent evaluation workflows to stay relevant as firms adopt these tools. Experiment with automated cross-referencing in n8n or Python to see how multiple prompts audit each other.

Source: Hacker News, The Decoder

Spot a mistake? Tell us on WhatsApp.