HM Revenue and Customs is building a fully digital tax and customs system, using AI to handle routine work connected to tax checks, fraud detection and customer support. The department is working with Microsoft and other tech companies as older systems move onto cloud platforms.
James Mitton, HMRC鈥檚 first Chief AI Officer, said, 鈥淭he aim is to become an agile department supported by a modern IT infrastructure, with innovation driving a better customer experience, helping us close the tax gap and become more efficient in how we operate.鈥
HMRC wants digital services that help people manage tax obligations without missing deadlines or entering incorrect information. Microsoft UK Stories reported that HMRC has already rolled out Microsoft 365 Copilot to 28,000 staff members, with another expansion planned for 50,000 workers.
Matt Vick, Head of Futures and Innovation in HMRC鈥檚 Strategy Function, said, 鈥淭he department has used AI tools and techniques like Machine Learning and Natural Language Processing for many years鈥 including compliance targeting, document analysis, debt risk prediction, and customer contact handling.鈥
Vick also said, 鈥淚t鈥檚 clear AI will be foundational to HMRC鈥檚 future.鈥
听
What Jobs Will AI Handle At HMRC?
听
One early project involves complaints handling. HMRC is building an AI agent that summarises complaints sent by taxpayers and helps advisers identify missing information before replying to customers.
Dave Johnson, one of the senior leaders working on HMRC鈥檚 digital programme, said AI systems will create 鈥渟marter, more intuitive customer journeys鈥, helping taxpayers avoid data entry mistakes, missed deadlines and accidental non compliance.
Johnson said, 鈥淭axpayers aren鈥檛 always aware of their obligations and unknowingly miss deadlines.鈥
听
More from Artificial Intelligence
- AI Is Now Better At Catching Fraudsters Than Human Analysts: Here鈥檚 Why That Matters
- World Environment Day 2026: How Tech Is Impacting Europe鈥檚 Climate Footprint
- Claude Is Surging Across Enterprise 鈥 Portal26 Just Made Governance Free
- How AI-First Agencies Are Reshaping UK Marketing In 2026
- Anthropic鈥檚 Claude Outage: Are Businesses Built Around AI Prepared For Failures?
- The Women Behind MENA鈥檚 Most Exciting AI Startups
- Why The Gulf Could Become The World鈥檚 Most Exciting Deployment Zone For Physical AI
- AI Experts React To Anthropic鈥檚 Plans To Publicly Release Mythos
AI will also help staff working on difficult tax cases involving huge collections of information and documents. Johnson said: 鈥淪ome of the more complex cases involve vast amounts of data and can take an awful lot of time and resource. Generative AI can summarise that casework information and streamline our compliance activities.鈥
HMRC said tax review decisions will stay under human supervision. Johnson said: 鈥淎ny decisions will be reviewed and signed off by a human in the process with the relevant tax expertise.鈥
听
How Will AI Help Deal With Fraud And Tax Errors?
听
Tech company Quantexa announced a 拢175 million agreement with HMRC lasting 10 years. The project focuses on fraud detection, tax compliance work and data analysis.
Quantexa said the programme will modernise HMRC鈥檚 data systems and help staff identify 鈥渢ax at risk more efficiently. The company also said the work will 鈥減rotect public funds鈥 and support 鈥済overned AI at national scale鈥.
The system uses Quantexa鈥檚 Decision Intelligence Platform to connect fragmented records from different HMRC databases. Quantexa said this creates 鈥渁 trusted, governed foundation for advanced analytics and the safe deployment of AI at scale to enable augmented and automated 诲别肠颈蝉颈辞苍-尘补办颈苍驳.鈥
Vishal Marria, founder and chief executive of Quantexa, said, 鈥淕overnments around the world are facing a common challenge: how to turn complex, fragmented data into confident, timely decisions.鈥
Marria also said, 鈥淏y creating context from data and embedding trusted, governed AI, we are helping HMRC improve how public sector organizations make confident, informed decisions. This is a blueprint for how the UK government deploys AI at scale.鈥