QE Tax Flash - HMRC's New AI Tax Tool

HMRC is mobilising AI to modernise its data infrastructure, supercharge fraud detection and spot errors in tax returns. HMRC’s systems are currently fragmented across legacy platforms, making it difficult to build a unified picture of fraud networks, connected entities or payment errors. The new platform under a £175m deal with British AI company Quantexa is expected to transform HMRC’s current manual investigation techniques by connecting internal HMRC records with external data sources to identify hidden networks of companies and individuals masking fraudulent activity. At the same time, HMRC has appointed its first Chief AI Officer, and is committed to rolling out Microsoft Copilot across its staff.

The tax gap stands at £46.8bn. The government has pledged 5,500 new compliance caseworkers and a compliance yield target of £50.4bn for 2025-26. HMRC’s new investment in AI is intended to achieve those yields in a way that increased headcount alone cannot. However, AI is not perfect and risks include hallucinations, lack of explainability of decision-making, and entrenchment of biases in underlying data.

Liesl Fichardt: “Taxpayers and advisers should ensure their records, structures, filings and disclosures are ready to withstand the increased scrutiny from HMRC when these tools are applied.”

Emily Au: “Taxpayers should be aware that HMRC's ability to identify discrepancies, trace connections and target investigations is about to improve materially, but the likelihood of errors derived from AI may also increase.”

Julius Konstantin Berling: “When a government will not raise taxes, it collects them harder. With headline rates largely fixed, scaling up compliance yield has become the primary revenue strategy – and a £175m AI investment is far from coincidental. Necessity, it seems, has become the mother of innovation.”