AI-assisted pilots with UK government organisations cut some legacy-system analysis from weeks to hours, including a test that took regression testing from around 15 days to 30 minutes. The results suggest AI could help make difficult-to-prioritise modernisation work more practical, though they come from a small set of varied systems.
Watch Desk analysis
What happened
The Government Digital Service worked with the Student Loans Company, the UK Hydrographic Office, the Met Office and delivery partner Kainos on pilots using AI to analyse and test legacy systems, according to UKAuthority’s account.
At the Student Loans Company, AI-assisted analysis of 33 applications written in Clojure took less than a day, compared with an estimated five weeks using traditional methods. At the UK Hydrographic Office, work on a 20-year-old desktop application increased test coverage from zero to 96 per cent. Regression testing that had taken around 15 days was reduced to about 30 minutes.
Why it matters
Old systems can be hard to modernise when their documentation is incomplete, dependencies are unclear and expertise is concentrated among a few people. Faster analysis could help teams see what they have, identify risks and plan upgrades before those systems become even harder to maintain.
The pilots are not a promise that every government system can be modernised at this speed. The authors caution that the systems and technical environments differed, and say the work still depended on experienced teams, good governance and access to the right information. AI can speed up the work; it cannot conjure up missing institutional knowledge by staring intently at a server.
Our read
The useful result is not simply a quicker test. It is that teams may be able to assess risky, neglected systems well enough to begin work that previously struggled to make the priority list. The reported examples are encouraging, but they are pilots, not a universal benchmark. Public bodies considering similar work should look for measurable results and keep engineers accountable for what the tools uncover.
What to watch
- Whether GDS publishes its playbook of good practice, which is still undergoing review.
- Whether more public-sector teams report comparable results on different legacy systems.
- Whether faster analysis leads to funded modernisation, rather than a quicker inventory of problems.
Discussion spark: Should government teams use AI to move faster on ageing systems, or wait for more evidence that the approach works across different technical environments?
Sources and evidence
- Source update (3 November 2026, 00:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.