hiw3c 日报精选
The Decoder 网页快照

KPMG的“指数项目”将AI代理与真实审计团队结合使用,并发现这些机器人在针对具体问题时能进行更深入的探究。

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

KPMG在对一家英国上市公司进行常规审计的同时部署了超过100个AI代理。据英国审计主管凯斯·伯内特称,这些代理在某些情况下提出的问题比审计团队更深入。英国会计监管机构FRC认为这种方式存在风险,因为AI代理分析的数据量远远超过传统抽样方法,从而可能揭示那些可作为高收益咨询业务推销的发现,并损害审计师的独立性。该文章《KPMG的“指数化项目”》将AI代理与真实审计团队结合使用,并发现这些机器人在个

来源:The Decoder · the-decoder.com

正文为英文,可一键机器翻译(仅首次需要等待)

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

KPMG deployed more than 100 AI agents across all phases of an experimental parallel audit of a UK-listed company for the first time. According to KPMG's UK head of audit, the agents dug deeper into individual questions than the regular audit team. A human had to step in when an AI agent tried to match regulations against company documents.

This year, accounting firm KPMG ran a full AI-powered parallel audit of a FTSE 250 company for the first time. According to a Financial Times report , the Big Four firm used more than 100 AI agents that worked through every phase of the annual audit alongside the conventional one. The official audit opinion was based on the regular audit.

In several cases, "AI was able to go so much deeper" than the regular audit team, said Cath Burnet, KPMG's UK head of audit, describing the results of the experiment, which KPMG internally calls "Project Exponential." In some situations, she said, the technology asked "better questions" than the firm's own auditors.

The agents independently collected, analyzed, and documented data and drafted audit conclusions. In one area, however, a human had to intervene because an AI agent couldn't correctly match a set of regulations against company documents.

Human oversight is still essential

KPMG says human involvement remained "critical." Staff set goals for the agents, provided context, reviewed work periodically, and challenged conclusions before the audit opinion was finalized. Two partners and a director worked alongside senior managers and junior staff.

The firm plans to expand the project to as many as 20 larger audits. The technology isn't yet available at the scale needed for thousands of audits, though, she said. Few companies currently have clean enough data for AI agents to work with reliably, she said.

KPMG says the system offers "full visibility and full traceability," recording agent interactions with documents down to individual characters and timestamps.

Regulators see an old problem in a new form

The UK's accounting regulator, the FRC, published guidance just days ago that applies existing independence rules to the use of AI. The core question is when auditors may share findings from AI-powered audits with their clients without compromising independence. Strict limits or outright bans apply to certain non-audit services for audit clients. A firm that advises a company may later end up reviewing its own recommendations, with an incentive not to upset the client.

AI agents intensify this conflict, according to the FRC, because they dig deep into a company's data during an audit. Where traditional auditors pull samples, AI agents can potentially analyze far more transactions. That process can surface findings that go beyond the audit itself, including inefficient processes, potential cost savings, and patterns that might never have come up in a conventional audit.

The FRC sees a risk that such findings could come across as advertising for extra services or be used as a sales pitch for consulting work. Under certain circumstances, the FRC says, sharing those findings could pose a "distinct threat" to auditor independence.

"Whatever technology a firm uses, the auditor remains accountable," said Mark Babington, the FRC's executive director of regulatory standards. Auditors must be able to explain how they used AI, challenge its results, and stand behind the evidence, he said. Certain findings, such as major weaknesses in internal controls, still must be reported to clients. The FRC says these mandatory disclosures don't count as non-audit services.

Auditor independence is a particularly sensitive issue in the UK. After construction and outsourcing group Carillion collapsed in 2018, a parliamentary inquiry found that KPMG, as the company's longtime auditor, had failed to challenge management's "increasingly fantastical figures" for 19 years. Two KPMG entities were fined a total of £21 million as a result.

The industry's business model under pressure

KPMG's experiment also raises a more fundamental question. If AI agents can take over the labor-intensive audit work of large teams of junior auditors, what does that mean for the industry's fee structure? KPMG sees "clear time-saving opportunities," according to the FT, but won't say whether that would lead to lower fees.

At the same time, KPMG International has already negotiated lower fees with its own auditor, according to an FT report, arguing that AI makes the work cheaper. Rival PwC has already run several "parallel audit" projects of its own.

An internal town hall at rival Deloitte in May showed the same pressure and upset the firm's own consultants. According to the Wall Street Journal, management presented a chart showing the share of hourly consulting revenue shrinking to a thin sliver of the growing overall market by 2035.

One Deloitte consultant summed up the message, saying, "They heavily implied our model is toast. We're basically getting replaced by robots."

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

Read on for the full picture. Subscribe for hype-free coverage.

Full access to every article on THE DECODER

Join the comments and community discussions

A weekly AI news recap via mail

6x/year: "AI Radar" — deep dives on the AI topics that matter most

Daily AI news, always up to date

Covered by a team with 10+ years in AI