Do you use AI in tax work?
For many, the answer to that is now yes, in some form anyway. It is something the 20:20 Innovation team have been speaking about for a while (see further details from Sharon Cooke following her 2025 Ethics and PCRT for Tax Advisers webinar) and there is now brand new guidance from the professional bodies.
Topical guidance on using Artificial Intelligence (AI) in tax work has been issued by the seven professional bodies responsible for producing the document 'Professional Conduct in Relation to Taxation' (PCRT). The new guidance is designed to be read in conjunction with existing PCRT guidance and discusses scenarios and possible safeguards allowing AI tools to be used ethically.
AI tools under the spotlight
The new Professional Conduct in Relation to Taxation (PCRT) guidance considers the increasing use of AI tools in the tax sector. Concerns about whether AI tools can be used responsibly may be developing as quickly as the technology.
- Rapidly developing technology is presenting opportunities for firms in both compliance and advisory services, as well as in conducting Client Due Diligence checks, mergers and acquisitions work, and carrying out technical research.
- Recent tax cases show that taxpayers, and even tax advisers, are being caught out after using AI to help with their legal arguments.
Who is the guidance aimed at?
Any tax professional who is using, or thinking about using, AI tools as part of their UK tax work will find the guidance relevant.
- It is not a technical guide on using AI, but provides topical guidance that complements the main PCRT guidance. Its particular aim is the ethical use of AI in tax work.
- The PCRT authors recommend using an AI specialist when someone wants to employ AI for their tax work but is unfamiliar with its use.
Interaction with the five fundamental principles
The topical guidance reviews the use of AI tools against the five fundamental principles that form the framework of PCRT. These five principles are:
- Integrity.
- Objectivity.
- Professional competence and due care.
- Confidentiality.
- Professional behaviour.
Using AI tools safely
Ethical risks, together with possible safeguards that can mitigate these risks, are discussed in the guidance. Some of these safeguards include:
- Disclosing the potential use of AI tools in engagement letters.
- If using AI is fundamental to delivering a piece of work, does the client need to be informed directly before work starts?
- Putting in place an AI usage policy that outlines acceptable use of AI tools and staff disclosure to senior colleagues when an AI tool has been used.
- How will an audit trail be maintained so that it can be demonstrated that reasonable care was taken?
- Avoiding the tendency to favour output generated from automated systems, even when human reasoning or contradictory information raises questions. Added caution is needed when using publicly available AI tools.
- Do you understand the limitations of the tool and which data sources it uses?
- Are data sources accurate, up-to-date and in line with current regulations, or might they include subjective opinions or biased information?
- Can you identify when an AI tool would not be suitable for a particular client or piece of work?
- Treating the output of an AI tool as if it has been prepared by a less experienced junior colleague, so that it is reviewed with appropriate scepticism.
- Can potential 'hallucinations', such as fictitious case law, be identified?
- Have principals and staff been provided with sufficient training so that they are competent in using an AI tool?
- Has data provided by a client or third party been generated by AI?
- Do you know how to ensure that tax return software does not automate the submission of a tax return?
- Maintaining confidentiality and complying with General Data Protection Regulation (GDPR) and Data Protection Act requirements.
- Is a data handling/AI usage policy needed?
- Can data input into publicly available AI models be sufficiently anonymised?
- How are requests from clients to not use AI tools handled?
- Understanding the limitations of an AI tool and mitigating the risk of irresponsible use.
- Are AI-generated responses reviewed to ensure they are worded in a professional manner and do not include inappropriate language?
Not sure what all the fuss is about?
Even if you do not use AI tools, it can still be useful to know what is available. As AI technology develops, you may find it being embedded into other software packages that you use.
- The guidance has a helpful appendix detailing the different types of AI tools.
Additional resources
Useful guides on this topic
