HMRC Signs £175m, Ten-Year AI Deal With British Firm Quantexa
HMRC has handed a £175m, ten-year contract to British AI company Quantexa, tasking it with improving fraud detection, reducing errors, and generally making the tax office less of a nightmare to deal with.
Quantexa's platform works by combining HMRC's internal data with external sources to surface hidden networks of companies and individuals involved in fraudulent activity. It will also help customer service staff handle queries more effectively, and chase down legitimate payments that have been filed under the wrong reference number. That last one sounds mundane but is apparently more common than you'd think.
The company's CEO Vishal Marria was quick to frame this as augmentation rather than automation. "In government environments, AI cannot operate as a black box. Decisions need to be transparent, auditable, and explainable," he told the BBC. Any AI-generated conclusions about taxpayers will still require human sign-off before anything happens.
On data security, Marria was equally emphatic: HMRC data stays within HMRC's environment, and staff working on the contract will be ring-fenced from the rest of Quantexa's business.
The backdrop here matters. HMRC complaints hit over 93,000 in 2024-25, up from 70,000 in 2020-21, according to a Freedom of Information request by the Contentious Tax Group. Slow response times are a recurring grievance. The department clearly needs fixing, and the government is betting AI can help.
Quantexa is valued at $2.6bn and counts HSBC and Vodafone among its clients. The choice of a British firm is deliberate. There's growing unease in Westminster about digital sovereignty, particularly after the £330m Palantir contract to build an NHS data platform raised questions about how much critical public infrastructure should be handed to American tech companies with opaque backgrounds.
Choosing a homegrown supplier doesn't automatically make the tech good, of course. But it does at least mean the government can credibly claim it's trying to keep sensitive data closer to home.
