The UK government banks £45bn a year on AI and digital efficiencies its own auditor says are not yet evidenced

The problem

The UK public sector is being asked to cut costs while its workforce keeps growing. The National Audit Office records that public sector employment has risen to 6.2 million, with staff costs of £260 billion, “yet departments face pressure to reduce costs while continuing to recruit for priority services” (source). Against that pressure, the government has pinned a large share of its savings plan on AI and digital technology (source).

What was built

This is not a vendor deployment; it is an independent audit of a claim. On 15 July 2026 the National Audit Office, the UK’s independent public-spending auditor, published “Government workforce planning – lessons learned” (HC 267, 2026-27), a lessons-learned report whose central theme is the impact of AI and digital technologies on the public-sector workforce (source). The report sets the government’s stated AI-and-digital efficiency expectation against the state of the plans meant to deliver it (source).

The outcome

The headline number is a government expectation, not a delivered saving. The NAO states that “Government expects substantial efficiencies from digital transformation and AI, amounting to £45 billion each year, while 76% of civil servants expect AI to change how they perform their jobs within the next five years” (source). The Register, reporting independently, records the same figure: “The NAO references the government’s claim that it can create efficiencies from digital transformation and AI amounting to £45 billion each year” (source).

What the auditor could not find is the working behind it. The NAO’s full report states that “the published efficiency plans do not provide details of how departments derived their expected workforce efficiencies” and warns “there is a risk that departments are not identifying and estimating workforce efficiencies consistently” (source). The report adds that “there is also a more fundamental risk that departments may not be identifying the full opportunities for efficiencies from reshaping workforces in response to wide-ranging digital and AI adoption” (source). The Register, reporting independently, carries the same finding verbatim (source).

The figure has also drawn parliamentary scrutiny. According to The Register, Parliament’s Science, Innovation and Technology Committee called the figure “worryingly optimistic,” describing it as an example of “hype” that could impede progress in digital transformation (source). That characterisation reaches this page through The Register’s reporting; the Committee’s own publication could not be fetched during verification, so it is recorded as a press-relayed secondary, not the primary Committee document (source).

How this was verified

Method: independent audit of the public record, no contact with any subject. The £45 billion figure, the 76% civil-servant figure, and the 6.2 million / £260 billion workforce context were taken verbatim from the NAO’s own press release for “Government workforce planning – lessons learned” (HC 267, 2026-27), published 15 July 2026 (Tier 1, independent auditor). The NAO’s “derived their expected workforce efficiencies” finding was confirmed verbatim from the primary NAO full report PDF (HC 267, 2026-27), fetched and archived this session (Tier 1), and independently corroborated verbatim from The Register’s 20 July 2026 report (Tier 2, independent press), which also carries the £45 billion figure. All three pages were fetched and archived to the Wayback Machine, and both the NAO full report PDF and summary PDF were saved to sources/. Date of verification: 24 August 2026.

Path to green

This is a claim-scrutiny story: the £45 billion is a government expectation that an independent auditor cautions is not yet evidenced, not a green delivery of a realised saving. The two critical claims, the £45 billion expectation and the NAO’s “derived efficiencies” finding, are now each carried by the primary NAO report HC 267 (Tier 1) and independently by The Register (Tier 2). The one remaining press-relayed element is the Science, Innovation and Technology Committee’s “worryingly optimistic” and “hype” characterisation, a supporting point whose primary is the Committee’s own publication on committees.parliament.uk, which blocked automated fetch (HTTP 403) this session. No subject is contacted to close any gap.

Related case files

• A US Inspector General audit of automated government benefit decisions, the same independent-auditor-versus-automation pattern
• A UK government machine-learning fraud model with a realised, verified saving, the contrast to an unproven efficiency claim
• A UK regulator’s scrutiny of AI training on member data, another case of independent oversight applied to a public AI claim

Sources

1. National Audit Office · “Effective government workforce planning is key to delivering productive, resilient and affordable public services” (press release for “Government workforce planning – lessons learned”, HC 267, 2026-27) · 15 July 2026 · Tier 1 · https://www.nao.org.uk/press-releases/effective-government-workforce-planning-is-key-to-delivering-productive-resilient-and-affordable-public-services/
2. National Audit Office · “Government workforce planning – lessons learned” (full report, HC 267, 2026-27) · 15 July 2026 · Tier 1 · https://www.nao.org.uk/wp-content/uploads/2026/07/government-workforce-planning-lessons-learned.pdf
3. The Register · “Auditors tell UK government to do the math before banking on £45B AI savings” · 20 July 2026 · Tier 2 · https://www.theregister.com/public-sector/2026/07/20/auditors-tell-uk-government-to-do-the-math-before-banking-on-45b-ai-savings/5274194