A Less Experienced Junior Colleague: Seven Professional Bodies on AI in UK Tax Work, January 2026

On January 19, 2026, the seven professional bodies that jointly prepare Professional Conduct in Relation to Taxation, known as PCRT, published guidance for their members on using AI tools in tax work. Their joint statement names them: the Association of Accounting Technicians, ACCA, the Association of Taxation Technicians, the Chartered Institute of Taxation, ICAEW, ICAS and the Society of Trust and Estate Practitioners. PCRT, it says, "sets out the principles and standards of behaviour that all members and students of the author bodies must adhere to when undertaking tax work."

The guidance is for any member or regulated firm using, or considering using, AI tools when advising on UK tax matters, in practice, in business or in the public sector. It says it "is not intended to be technical guidance on the use of AI," and is meant "to complement the main PCRT guidance." "If a member fails to adhere to the principles set out in PCRT they are liable to be subject to the disciplinary process." Parts of it follow, not the whole.

Responsible for Any Work

"Members are ultimately responsible for any work they produce, and for regulated firms any work which the firm prepares, irrespective of the use of AI tools in its creation." Outputs from AI tools, it says, should not be used as authoritative tax or legal advice, and a qualified professional should review them in the context of the specific client.

"The output from an AI tool should also be regarded as if it were prepared by a less experienced junior colleague and reviewed with appropriate scepticism." It calls hallucinations the cases where a tool "produces data which is nonsensical and/or inaccurate, but which is presented as factual in the response." Confirming that case law or legislation cited in an output exists can help, it says, and it points to a case it cites as [2023] UKFTT 1007 (TC).

One of its examples is tax return software that can automate the preparation and submission of a tax filing and "can be configured to make human review optional." Beside it, the guidance says to "Obtain an understanding of the workflows that can be automated and the options for ensuring that the data and the submission can be subject to an appropriate review."

Telling the Client

"Transparency is an important element of ethical AI usage." A statement in the engagement letter that AI-enabled software may be used can help, the guidance says, and consideration should be given to telling the client about any actual use of AI tools when the work is delivered. Where AI is fundamental to the deliverable, "it may be appropriate to inform the client directly prior to commencement of the work."

When a client queries the work, members "should be able to explain how the conclusion was arrived at, even if the work was generated by AI." Justifying the tool's conclusions, it says, may include outlining any assumptions made in generating the results, or the prompts given to the tool. Where relevant, members should consider keeping copies of web pages and screenshots of prompts "as part of their audit trail on record," as evidence that reasonable care was taken.

Bias and the Tool

The guidance defines automation bias as "a tendency to favour output generated from automated systems," even when, in its words, "human reasoning or contradictory information raises questions as to whether such output is reliable or fit for purpose." "AI tools are designed to create a response to satisfy the request rather than constructing an unbiased and fair output."

It gives the case of a tool built for one kind of client being used for clients whose circumstances differ, and says that using a tool to save time "should not be at the expense of an increased exposure to bias," or of relying on data that does not apply. And a new version of a familiar tool should be approached with caution "until the operational effectiveness of the tool has been demonstrated to be at least of the same standard as the earlier version."

The Client's Data

"The input of client data into publicly available AI tools is likely to constitute a breach of client confidentiality, unless the client has consented to this." Once information goes into such a tool, "control over that data is relinquished." Data put into public AI models should be anonymized and generic, and members should take care that a client cannot be identified by other means, such as an uncommon service the client is known to provide.

It also covers a client who asks that AI tools not be used on their affairs. The member "should acknowledge the client's request and may need to seek specialist advice," and may consider discussing with the client what they would regard as an unacceptable use. Elsewhere it notes that software packages "often have AI tools (or similar) embedded within them."

That leaves a practical question for a firm whose tax software already has AI built in. What exactly would it be promising a client who asks for no AI at all?