Watchdog sees AI prize for UK government finance teams

TL;DR:

  • The National Audit Office says AI presents “a significant opportunity” to transform the Government Finance Function and its 9,000-plus members.
  • It urges structured experimentation and a consistent cross-government approach, but flags legacy IT and poor-quality cost data as barriers.
  • Departments have signed up to deliver around £14bn in annual efficiencies under the 2025 Spending Review.

The UK’s spending watchdog has urged the government to push harder on AI in its finance teams, while warning that ageing systems and unreliable data could blunt the gains. A new National Audit Office report finds the Government Finance Function (GFF) has made “an excellent start” on its 2030 strategy, but says future progress depends on overcoming “enduring structural and cultural barriers”.

Opportunity, with caveats

The NAO wants the GFF to lead on testing and adopting AI for finance tasks, using its innovation committee to define “priority problem areas” where the technology could lift productivity and quality, then sharing what works so departments can adopt with confidence. The report describes the pace and potential of AI in government as “a significant opportunity” — but is candid about the obstacles: legacy IT, the poor quality of cost data across government, and weak incentives for non-finance staff to prioritise sound financial management.

NAO head Gareth Davies tied the case to fiscal pressure, arguing strong financial management is “critical to helping departments make better decisions, improve productivity and deliver better value for taxpayers”. Public Accounts Committee chair Sir Geoffrey Clifton-Brown was sharper, warning the benefits will only materialise “if government prioritises the right specialist skills and tackles the longstanding issues of poor-quality cost data and legacy IT systems”. The data-quality caveat is the crux: AI applied to unreliable cost figures will produce confident but misleading outputs.

Looking forward

The report sits within a broader push to wire AI into UK public services, from the DVLA’s AI phone system halving call navigation times to NHS adoption sandboxes. The difference here is scale and sensitivity — public money. With departments committed to roughly £14bn in annual efficiencies, the temptation to claim quick AI wins will be strong. The NAO’s message is a useful corrective: foundations first. Standardised financial data and modern systems are unglamorous, but without them the productivity prize stays theoretical.

Share this article