Auditors tell UK government to do the math before banking on £45B AI savings
The UK's National Audit Office (NAO) has urged civil service leaders to properly assess how artificial intelligence will affect the size, shape and skills of the public sector workforce before assuming it will deliver the government's projected £45 billion in annual savings. The watchdog says departments have yet to show how AI-driven efficiencies would translate into changes in staff numbers, roles, grades and required capabilities, warning that without this planning the government risks overestimating or unevenly calculating the benefits of digital transformation.
The NAO's report notes that expected productivity gains, such as digital self-service for tax and benefits, and automated checks on planning applications, could reduce staffing needs in some areas, but the public sector currently lacks the digital skills required to realise them. The £45 billion efficiency figure has already drawn scepticism, with the Institute for Government calling it a "huge amount" given most spending goes on salaries and infrastructure, and Parliament's Science, Innovation and Technology Committee branding it "worryingly optimistic" hype. The report also cites Cabinet Secretary Dame Antonia Romeo's own admission that workforce plans do not adequately reflect coming AI-driven change, and flags potential ethical concerns that could limit AI deployment in some areas.
In short
- NAO says government hasn't planned how AI will reshape civil service jobs
- £45bn annual AI savings claim already called "worryingly optimistic" by MPs
- Public sector currently lacks digital skills needed to realise AI efficiencies
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
The UK government faces significant fiscal pressures, and AI genuinely does offer potential for automating routine administrative tasks and improving service delivery across health, welfare, and planning systems. Whilst the £45 billion target is ambitious, bold goals are necessary to drive innovation and efficiency in the public sector; critics who counsel caution risk becoming prisoners of the perfect being enemy of the good. Phased implementation with iterative learning can address the skills gaps and planning concerns flagged by the NAO, and waiting for ideal conditions may mean squandering an opportunity to modernise a sector that has long lagged behind private industry.
The case against
The National Audit Office's warning reflects a serious problem: the government has published a savings target without demonstrating how it would translate into concrete staffing and structural changes, and most public sector spending goes to salaries and essential infrastructure, not discretionary expenses amenable to efficiency gains. The public sector currently lacks the digital skills needed to implement such transformation, and rushing into AI deployment without proper workforce planning risks either failing to achieve the promised savings or causing significant disruption to public services. The government should proceed cautiously, doing the hard, unglamorous work of auditing which roles can genuinely be eliminated or transformed before making financial commitments based on current projections.
Explained simply
The UK government has promised to save £45 billion a year through using artificial intelligence to make the public sector more efficient. The National Audit Office (NAO), an independent watchdog that scrutinises government spending, has questioned whether this figure is realistic because departments have not properly planned how AI would actually change the workforce.
The NAO's main worry is that government departments have not shown how AI-driven efficiency gains would translate into changes to staff numbers, roles and the skills workers would need. Without this detailed planning, the watchdog warns the government risks either overestimating the savings or delivering them unevenly across the civil service.
The £45 billion figure has already faced scepticism from other observers, including the Institute for Government and Parliament's Science, Innovation and Technology Committee, who say it seems unrealistically high. Additionally, the public sector currently lacks many of the digital skills needed to properly implement AI in the first place, further suggesting the savings may not materialise as planned.
AI Art Culture Government Politics Technology UK World