KPMG's "Project Exponential" put AI agents alongside a real audit team and found the bots dug deeper on individual questions

KPMG placed AI agents alongside a real audit team in an experiment called Project Exponential. The bots dug deeper on individual questions than the human auditors.

The Big Four firm ran the test inside a working audit engagement, not a lab. The results show AI agents can handle specific, detailed audit tasks while humans stay in control of judgment and sign-off.

Do not expect a fully autonomous audit anytime soon. The test was designed to find where AI adds value and where it does not.

Project Exponential paired AI agents with human auditors

KPMG built Project Exponential to see how AI agents perform in real professional work. The firm put multiple agents alongside an actual audit team and let them work through the same individual questions.

The setup mirrored a live audit. Each side reviewed documents, extracted insights, and flagged issues for follow-up.

The experiment measured depth of analysis, not just speed.

The bots dug deeper on individual questions

The central finding: AI agents went further on individual questions. They examined more underlying documents and followed more threads on each specific item.

Human auditors often rely on experience to decide where to dig. The AI agents methodically investigated every question with consistent depth.

The key insight: AI agents do not just answer questions. They unpack them, layer by layer, until the detail is exposed.

That depth is valuable because audit risk hides in small details.

AI agents do not get tired. They do not skim. They apply the same rigor to the first question and the hundredth.

Human judgment still leads the audit

KPMG kept human auditors in charge of the process. The AI agents produced findings, but people reviewed them before anything entered the audit file.

Auditing standards require professional judgment. A bot cannot sign an audit opinion or take regulatory accountability.

The audit file remains the responsibility of named professionals. Regulators will not accept a bot’s conclusion as sufficient evidence.

This division of labor is the model the industry is moving toward.

Why this matters for the audit profession

Audit teams face pressure to do more with the same headcount. AI agents can extend their reach on every question.

  • AI agents increase audit depth: Bots examined each question more thoroughly than the human team.
  • Humans retain final control: Auditors reviewed bot findings and made the final calls.
  • The hybrid team is the future: AI handles volume and depth while humans handle judgment.

Clients expect more scrutiny of financial records. Deeper questioning means fewer missed risks. That is why the KPMG result matters beyond one firm.

The bigger picture for professional services

KPMG has pushed AI into multiple service lines. Project Exponential stands out because it tested AI agents in a live workflow, not in a demo.

The results suggest AI agents can act as collaborative team members. They take an individual question and pursue it well beyond the surface.

This points to a future where professionals focus on interpretation while agents handle exhaustive investigation.

Other professional services firms are watching. If AI agents can dig deeper on audit questions, they can likely do the same for tax, consulting, and risk work.

The bottom line

Project Exponential was not about cutting staff. It was about augmenting the people already doing the work.

The experiment produced a clear answer: AI agents dig deeper, and humans still decide. Both sides matter.

The audit of the future will not be purely human or purely automated. It will be a partnership.

Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.

What are your thoughts on this? I’d love to hear about your own experiences in the comments below.