在里程碑式审计试验中,AI代理比毕马威会计师提出更好的问题
AI agents ask better questions than KPMG accountants in audit trial

原始链接: https://europeanbusinessmagazine.com/ai-agents-better-questions-kpmg/

KPMG英国在“Project Exponential”项目中,对一家富时250指数公司账目测试了100多个AI代理,并与人工审计员协同工作。这些代理覆盖了审计工作流程,在某些情况下比人类更深入地探查问题并提出人类遗漏的问题。然而,这并非正式法定审计;人工监督仍必不可少,包括在代理错误应用监管要求时进行干预。KPMG计划扩展至最多20个更大型审计,但表示技术尚未准备好应对数千个审计,数据质量仍是障碍。该实验凸显了AI在自动化数据收集、分析和文档记录方面的潜力,而资深专业人士专注于判断,这可能改变审计经济学和四大运营模式。监管机构警告AI必须维护审计师的独立性、怀疑态度和准确性;最终责任仍由人工审计员承担。关键问题是机器最终将执行多少审计工作,人类又将扮演什么角色。

KPMG 在审计中测试了 AI 代理,发现它们有时提出比人类审计员更深入、更尖锐的问题,但至少一个代理在将监管要求与公司文件匹配时出错。Hacker News 评论者普遍认为,大语言模型可用于详细、清单密集的工作并减少人类疲劳,但争论审计是否真正公式化。一些人认为,四大审计可能受客户激励、成本压力和初级人员配置影响,而另一些人强调人类审计员在重要性和背景上运用判断。担忧包括 AI 过度生成、同质化风险、被操纵或利用的可能,以及人类问责的必要性。支持者反驳称,AI 可扩展穷尽式分析并一致执行程序。总体而言,AI 作为审计辅助工具前景可期,但不能完全替代,监督、激励、准确性以及需要多少人类审查等问题仍未解决。
相关文章

原文

LONDON, November 18, 2026 — Artificial intelligence has taken another significant step into the world of professional services after KPMG revealed that AI agents examining the accounts of a major UK-listed company sometimes investigated financial issues more deeply and asked better questions than its human audit team.

The experiment, conducted by KPMG UK using more than 100 AI agents, marks an important development in the application of artificial intelligence to one of the most labour-intensive areas of the financial services industry. It also raises an uncomfortable question for the accountancy profession: if machines can increasingly perform the analytical work traditionally undertaken by highly trained professionals, how quickly will the economics of auditing change?

According to the Financial Times, the project, codenamed Project Exponential, examined the financial statements of an unnamed FTSE 250 company. The AI agents worked alongside a conventional audit team, undertaking tasks across the audit process rather than simply assisting with isolated administrative functions.

Join The European Business Briefing

New subscribers this quarter are entered into a draw to win a Rolex Submariner. Join 40,000+ founders, investors and executives who read EBM every day.

Subscribe

KPMG UK head of audit Cath Burnet said the technology had, in several instances, gone considerably deeper into issues than the regular audit team, identifying areas requiring further investigation and asking questions that human auditors had not necessarily raised.

The findings suggest that AI’s potential in professional services may extend well beyond improving productivity. It could fundamentally change how complex financial information is examined, interpreted and challenged.

From supporting accountants to conducting audits

Project Exponential represents a shift from using AI as a supporting tool towards deploying autonomous agents across an entire audit workflow.

The agents collected information, analysed and documented data, developed draft conclusions and observations, and suggested subsequent steps. They were also able to work alongside one another, effectively dividing responsibilities and coordinating elements of the investigation.

The technology was developed over approximately four months. However, KPMG was careful to distinguish the experiment from a formally approved statutory audit. Although an audit partner signed off on the experimental exercise, its findings were not used as the basis for formally approving the company’s accounts.

Human oversight remained essential. During the trial, an auditor had to intervene when one AI agent failed to correctly match regulatory requirements with the company’s documentation. The episode illustrates both the capabilities and limitations of current AI systems when dealing with complex regulatory frameworks.

KPMG plans to expand the initiative to as many as 20 larger audits, although the firm acknowledges that the technology is not yet sufficiently mature to be deployed across thousands of engagements. Data quality also remains a significant obstacle, with some businesses lacking the clean, structured information required for AI agents to operate effectively.

A challenge to the Big Four business model?

The implications for the accountancy industry could be substantial.

Traditional auditing relies on large teams of junior accountants undertaking detailed and repetitive work, supervised by experienced managers and partners. AI agents potentially offer a different operating model, automating significant portions of data collection, analysis and documentation while allowing senior professionals to concentrate on complex judgements and exceptions.

For firms such as KPMG, Deloitte, PwC and EY, the attraction is clear: faster audits, potentially lower operating costs and the ability to analyse much larger volumes of financial information.

However, whether those savings will ultimately be passed on to clients remains uncertain. KPMG has acknowledged the potential for significant time savings but has not confirmed whether AI-driven efficiencies will translate into lower audit fees.

Human judgement remains the final safeguard

The transition also presents regulatory and professional challenges. The UK’s Financial Reporting Council has warned that the use of AI must not undermine auditor independence, integrity or professional scepticism.

There are also concerns about data security, over-reliance on automated conclusions and the possibility of AI generating inaccurate or fabricated information. The fundamental responsibility for an audit opinion remains with the human auditor, regardless of the technology used.

The KPMG experiment does not establish that AI can replace accountants or independently deliver a reliable statutory audit. It does, however, demonstrate that AI agents are beginning to challenge professionals in areas that were once considered dependent on human expertise.

For the accountancy profession, the question is no longer simply whether AI will become part of the audit process. It is how much of that process will eventually be performed by machines — and what role human accountants will play when it does.

联系我们 contact @ memedata.com