Judgement Still Matters

Artificial intelligence is becoming part of everyday tax and accounting work. It can summarise legislation, draft arguments, and process large volumes of data quickly. But recent cases are drawing an important line: AI can support the work, but it cannot replace professional judgement.

That point came through clearly in Elden v HMRC [2026], which concerned disputed appeals against HMRC closure notices regarding income tax and capital gains tax, amounting to around £210,000. However, the case was not really about the underlying tax dispute: it became about how the appeal was being handled. HMRC raised concerns about a skeleton argument submitted, which cited five cases that were either incorrectly or misleadingly summarised and irrelevant, suggesting artificial intelligence may have been used. The tribunal found that case summaries were inaccurate, authorities had been mischaracterised and the material had not been properly checked. The issue was not the use of AI itself. The issue was the lack of care taken in relying on its output. The tribunal’s response was telling. It imposed stricter requirements for future submissions, including full copies of authorities, direct quotations and precise references.

This is not an isolated issue. In Mata v Avianca [2023], a US court sanctioned lawyers who relied on AI-generated cases that did not exist. Roberto Mata sued Avianca Airlines for injury, but his attorneys used AI, which ‘hallucinated’ non-existent precedents complete with fake quotes and citations, which were submitted to the court. In the UK, the Solicitors Regulation Authority has also made clear that while AI can be used in professional work, accountability for the final product does not shift to the tool.

The risk is not only that AI produces weak technical analysis. In a tax context, that can mean poorly supported positions, flawed submissions or increased enquiry exposure.

That principle matters just as much in tax as it does in wider financial reporting. Where analysis feeds provisions, disclosures, uncertain positions, or board reporting, it can quickly become an accounting and governance issue. A weak technical conclusion can flow into the financial statements, the audit process and the credibility of the wider finance function.

That is why the real issue is not speed, but control. AI cannot assess nuance, apply scepticism, weigh competing interpretations or take responsibility for a filing, a disclosure or a legal submission. In both tax and accounting, the core disciplines remain the same: evidence, judgement, review and clear ownership.

The lesson is not to avoid AI. It is to use it properly. It should be treated as a starting point, not a conclusion. As tax compliance and financial reporting become more data-led, more transparent, and more heavily scrutinised, experienced advisers remain essential. Their value lies not simply in producing information, but in interpreting it, challenging and standing behind it.