HM Revenue and Customs is not beginning its AI strategy with a chatbot placed over decades of fragmented systems. The UK tax authority has selected SAP to modernise the platform beneath tax administration, creating a managed and sovereign foundation for later automation.
The Enterprise Tax Management Platform supports more than 45 tax regimes and manages over £800 billion in annual revenue. Tens of thousands of employees depend on it. Moving such a system is therefore a national-infrastructure project before it is an AI project.
HMRC plans to use RISE with SAP, SAP Business Technology Platform, and AI capabilities in a UK Sovereign Cloud. The stated goals include simpler architecture, better analytical access, faster insights, and improved taxpayer interactions.
Unified Data Comes Before Automated Decisions
Machine learning needs consistent information across transactions, taxpayers, cases, and time. Fragmented on-premises systems make that difficult because definitions, interfaces, and update cycles differ.
Modernising the core provides an opportunity to establish authoritative records and reusable data services. Employees can access analysis through a more coherent interface, while automated processes can operate on the same governed information.
This does not mean every historical record should be combined without limits. Tax data is highly sensitive, and access must reflect purpose, role, and legal authority. A unified platform should improve control over who can see and use information, not simply make it more available.
Data quality also becomes visible during migration. Duplicate identities, inconsistent fields, and undocumented process rules need resolution before they are fed into automated decisions.
The migration plan should measure continuity, reconciliation accuracy, user performance, and taxpayer outcomes—not only the number of systems retired.
Sovereignty Is an Operating Requirement
HMRC will host the new capabilities in SAP’s UK Sovereign Cloud to meet local data-residency, security, and compliance needs. Sovereignty is particularly important when a commercial platform supports tax records and decisions at national scale.
Location alone is insufficient. HMRC needs clear control over administrators, encryption keys, operational access, logging, updates, incident response, and the conditions under which suppliers can handle data.
The platform must also remain resilient as demand changes. Tax deadlines create sharp peaks, while policy changes can require new rules and forms quickly. Managed cloud infrastructure can provide flexibility, but essential services need tested recovery arrangements and clear public accountability.
SAP describes the project as strengthening resilience, security, and sustainability. Those outcomes should be evidenced through availability, recovery tests, energy use, vulnerability response, and reduction of unsupported legacy components.
Tax AI Requires Explainable Public Decisions
HMRC and SAP plan to define AI capabilities aimed at taxpayer experience and decision support. Potential value lies in faster case analysis, more responsive reporting, and reduced friction in interactions.
The risk depends on the action. An assistant that helps an employee find guidance requires different controls from a model that prioritises an investigation or influences an individual tax outcome.
Higher-consequence uses need traceable sources, evaluation across taxpayer groups, human review, appeal routes, and records showing how the result was produced. Public-sector efficiency cannot come at the cost of procedural fairness.
Employees should be involved in workflow design because they understand the exceptions and informal checks that keep a complex tax system functioning. Automation that ignores those practices can move errors more quickly.
HMRC’s approach recognises that technical debt and AI ambition are linked. A model cannot create a coherent tax service while the information and processes beneath it remain fragmented.
The programme will be judged over years, not through a demonstration. Its success would look like a more resilient platform, fewer manual handoffs, better decisions, and simpler taxpayer experiences. AI may support those results, but the difficult achievement is the modern public infrastructure that makes them possible.
