A contract review that once took hours now has a first draft in five minutes. The accountant’s work is changing along with it.
In Brief
AI has begun reading contracts and drafting audit working papers. Less time spent writing is welcome, but it also means that tools are taking on work people used to do. If auditors remain responsible, how will their roles and staffing needs change?
1. Contract review and working paper preparation
Give AI several complex contracts that would take an audit in-charge hours to review. About five minutes later, it produces a draft in working paper form. I have seen this in practice. The source documents and conclusions still need to be checked, but the work no longer has to begin with the first sentence on a blank page.
Within Big Four firms, tools such as ChatGPT 6.0 Astra and Claude Opus 5.5 are being used through internal systems designed to protect client information. AI is moving from something people consider trying to something they routinely open when they start work.
Public announcements point in the same direction. In May 2026, PwC and Anthropic said that Claude was already available through PwC’s internal tool, ChatPwC. That month, KPMG also announced plans to provide access to Claude across its global workforce. The specific models and extent of use vary by firm and working environment. Anthropic, PwC alliance announcement, Anthropic, KPMG alliance announcement
A tool that answered questions about standards and refined sentences now produces the document to be reviewed. AI is beginning to perform much of the work accountants did when they read documents, organised the information and turned it into a working paper.
2. Task automation and changing roles
For practitioners, this is welcome. Deadlines are fixed, and there is plenty to review. Less time spent organising information and preparing working papers makes it easier to get on with the remaining work. The benefit has been clear in my own experience.
But could what we call efficiency also be evidence that AI is replacing part of the accountant’s work?
AI is now doing tasks that people used to perform. For those tasks, replacement has already begun. The entire profession does not have to disappear before we can say that some work is being replaced. That does not mean staffing will immediately fall in proportion to the time saved on drafting. Teams may use the time for deeper review or take on more work with the same people. Time spent preparing working papers and total audit hours need to be considered separately.
What is clear is that the basis for allocating work may change. If AI takes over the draft that a new associate once prepared for the in-charge to review, the team must reconsider what each level does and where it spends its time.
3. Auditor responsibility and staffing
Auditors must still obtain audit evidence to support their judgements when they use AI-generated working papers. A change in the tool used to prepare a document does not reduce the auditor’s responsibility.
US PCAOB auditing standards and UK FRC guidance help explain this point. The table summarises standards and guidance from their respective jurisdictions; it is not a single set of rules that applies to every audit worldwide.
| Standard | Main Point | IMplication for AI-Generated working papers |
|---|---|---|
| PCAOB AS 1000,.10-13 | Support significant judgements and conclusions with sufficient appropriate audit evidence. Exercise professional scepticism and judgement. | Adopting an AI conclusion still requires the auditor’s judgement and evaluation of evidence. |
| PCAOB AS 1215,.06-06A | Record procedures, evidence, conclusions, and who performed and reviewed the work, so an experienced auditor with no prior connection to the audit can understand it. | Polished writing is not enough. The work actually performed and the basis for the conclusion must be clear. |
| FRC, Generative and Agentic AI Guidance, March 2026 | Explains review and controls suited to the purpose and risks of AI output. Professional judgement determines the nature and extent of review | Managing audit quality includes how AI is used and how its output is reviewed. |
When publishing its guidance, the FRC stated that the regulatory responsibilities of audit firms and those responsible for AI adoption and audit quality remain unchanged. The guidance explains how to use AI within those existing responsibilities. FRC announcement, March 30, 2026
The need for someone to take responsibility does not, by itself, establish a need for the same number of people. As AI performs more tasks, the work left to people and the way it is allocated may change. In considering the future of accounting, we need to look at both who holds ultimate responsibility and what work is required to fulfil it.
4. AI use and review capability
In practice, reliance on AI is growing from engagement leaders to new associates, to the point where it can feel difficult to make progress without it. Becoming accustomed to a useful tool is natural. It is still worth considering what we have come to depend on it for.
Working more slowly without AI is different from being unable to judge whether its conclusion is sound.
A well-presented working paper can make the review feel substantially complete. Yet a missing contract term or an assumption that does not fit the company’s circumstances could change the conclusion. Reviewers need to recognise those omissions and assumptions before taking comfort from polished writing.
The FRC also addresses automation bias: the risk of placing too much confidence in AI output. It explains the need to assess accuracy and completeness, as well as consistency with the reviewer’s understanding of the entity. The ability to prepare a working paper quickly with AI needs to be accompanied by an understanding of what to check. FRC, Generative and Agentic AI Guidance, p. 37
5. Judgement and practical experience
That judgement has traditionally developed through reading source material, preparing working papers and receiving review comments. As preparation changes, learning must change too. The next step is to examine the judgements experienced accountants are already adding to AI output.
Key Takeaways
- AI produces working paper drafts for accountants to review. Producing a draft and completing the work after review are distinct stages.
- Work allocation and staffing needs can change even while auditors retain their responsibilities.
- As the ability to assess AI output becomes more important, work must continue to provide the experience through which that judgement develops.
What, then, should accountants do to avoid being replaced by AI?
The next article begins with that question.
References
- Anthropic, PwC is deploying Claude to build technology, execute deals, and reinvent enterprise functions for clients, May 14, 2026.
- Company annual report, FY25, note 4.Anthropic, KPMG integrates Claude across its core business and workforce of more than 276,000 in strategic alliance, May 19, 2026.Company annual report, FY25, note 4.
- Public Company Accounting Oversight Board (PCAOB), AS 1000 — General Responsibilities of the Auditor in Conducting an Audit, paragraphs .10–.13.
- Public Company Accounting Oversight Board (PCAOB), AS 1215 — Audit Documentation, paragraphs .06–.06A.
- Financial Reporting Council (FRC), Generative and Agentic AI Guidance, March 2026, pp. 37–38.
- Financial Reporting Council (FRC), Innovative new guidance supports audit firm adoption of emerging AI technologies, March 30, 2026.