The UK government has reported significant productivity gains following a wide-scale trial of AI coding assistants from Microsoft, GitHub, and Google. The trial, which ran between November 2024 and February 2025, involved more than 1,000 government developers across 50 departments.
Results indicate that AI tools such as GitHub Copilot and Google Gemini Code Assist helped developers save an average of one hour per day—equivalent to around 28 working days over a year. Officials have positioned this as a major step toward making public sector operations more agile and modernized.
Results of the AI Trial
According to data released from the review, the biggest time savings came from tools that supported developers in:
- Writing first drafts of code
- Reviewing and debugging existing code
- Assisting in problem-solving and task completion
Although only about 15% of AI-generated code was used without human edits, engineers reported significant reductions in repetitive tasks. The trial showed that 72% of developers believed the tools provided good value, while 65% said tasks were completed faster and 56% reported problem-solving improvements.
Notably, 58% of participants expressed that they would prefer not to return to working without AI coding tools, signaling a strong appetite for permanent integration into workflows.
Government Endorsement
Technology minister Kanishka Narayan praised the initiative, noting that the public sector must catch up with technological innovation.
He emphasized that the findings reveal how hungry engineers are to adopt AI responsibly and effectively to accelerate the delivery of critical digital services.
Google Cloud’s Tara Brady welcomed the outcome, calling it a landmark moment for AI adoption in the UK public sector. She described the collaboration as proof of how partnerships between government and cloud providers can drive transformation in essential services.
The government has also set ambitious goals, with Prime Minister Keir Starmer expressing plans to expand AI implementation further, ultimately aiming to save more than £45 billion for taxpayers through AI-driven efficiencies.
Concerns About AI-Generated Code
Despite the positive results, industry experts have pointed out challenges. Martin Reynolds, CTO at Harness, highlighted that 85% of AI-generated code still required manual editing before reaching later stages of production such as security testing, deployment, and verification. He argued that while AI boosts initial coding speed, remediation responsibilities remain a significant burden.
The quality and security of AI-produced code has also come under scrutiny. Cloudsmith’s Nigel Douglas warned that without strong security-aware practices, AI coding assistants could inadvertently introduce vulnerabilities into sensitive government software systems. He stressed the growing need for secure-by-design tools and stronger policy enforcement in AI deployment.
A recent study further underscored this concern, showing that developers often spend additional time fixing AI-generated outputs, which can undermine the very efficiency gains the technology promises.
The Path Ahead
The UK government’s trial demonstrates both the potential and pitfalls of AI in software development. While AI coding assistants have already proven capable of saving significant time and improving task completion, challenges around quality assurance, remediation, and security must be addressed.
As adoption expands across public sector services, the focus will need to shift toward balancing efficiency gains with robust safeguards. For now, however, the trial confirms that AI-powered coding is no longer an experiment—it is fast becoming a core enabler of government digital transformation.
Read more such articles from our Newsletter here.
