The UK government has been telling anyone who will listen that artificial intelligence could save its public sector £45 billion a year. The National Audit Office, the country’s spending watchdog, has a polite but firm response: prove it.

A July 20 report from the NAO calls on ministers to rigorously verify that headline number before baking it into workforce plans, budgets, or strategic blueprints. The figure originated from the Department for Science, Innovation and Technology’s “State of digital government review” published in January 2025, which estimated that digital transformation and AI could unlock efficiencies worth 4-7% of total public sector spending.

The gap between ambition and evidence

The Science, Innovation and Technology Committee labeled the estimate “worryingly optimistic” back in June 2026.

Independent analysis paints a considerably more modest picture. The Institute for Government and the Ada Lovelace Institute have each produced alternative estimates ranging from £23.6 billion to £24 billion annually. Still substantial numbers, but roughly half the government’s projection.

The NAO report doesn’t just question the topline figure. It highlights a deeper structural problem: government departments haven’t done the homework needed to translate AI ambitions into actual workforce transition plans. The watchdog specifically flagged that departments need to integrate digital transformation impacts into their workforce planning.

The workforce question nobody wants to answer

The NAO found that detailed workforce efficiency plans tied to AI adoption are largely absent across departments.

The ethical dimension matters too. Automated decision-making in government services, from benefits processing to healthcare triage, carries risks that private sector deployments don’t. Getting an AI recommendation wrong at a government agency can affect someone’s housing, healthcare, or legal status.