from the Edge 27/03/26
RPNA Intelligence Briefing
Public Sector AI: Fortnightly Briefing
27 March 2026
Lead Story
GOV.UK Chat opens to app users after pilots hit 90% accuracy
GDS has completed two public pilots of GOV.UK Chat, during which more than 10,000 users asked 26,000 questions about government services. Accuracy rose from 76% at the outset to 90% by the end of the trials, all 508 jailbreak attempts were blocked, and 73% of users rated the service as useful. Access is now being extended to all GOV.UK app users, with a wider rollout across the GOV.UK website planned for later in 2026.
Analysis
From pilots to practice: why most public sector AI is going nowhere
Leadership
Sixth out of ten is not a crisis, but it is a signal. The UK has proof of concept. What it lacks is the accountability structure to move from "we ran a pilot" to "this is how we work now." Someone has to own scale, and right now that job description does not exist in most organisations.
Governance
The EU delay buys time, but time without direction is not a gift. UK public bodies now face a choice: wait for the regulatory picture to clear, or build governance that will hold up regardless. The organisations that use this window well will not be the ones who waited.
Workforce & Skills
Sixty per cent say AI is easy to use. Sixty-three per cent admit they barely understand it. The gap runs deeper than training. Organisations are counting familiarity as capability and building strategy on top of it. That is where the risk sits.
UK
MHCLG launches Local AI programme for councils
A new team within the Ministry of Housing, Communities and Local Government is working directly with councils to support responsible AI adoption across local government. The programme is focused on building and sharing reusable AI tools, starting with a transcription service for public-facing workers, and strengthening data and governance foundations. This is the first time MHCLG has dedicated a team specifically to AI in local government.
MHCLG funds £1.1m pilot to use data to reach vulnerable residents earlier
The government is funding pilots in Greater Manchester and Bournemouth, Christchurch and Poole to test data standards that let councils identify vulnerable residents before a crisis develops. The pilots use the SAVVI and Open Referral UK data standards to enable councils to share and act on vulnerability signals across services, a step towards proactive rather than reactive care.
House of Lords Science and Technology Committee opens inquiry into NHS AI and personalised medicine
The committee launched an inquiry into why the NHS struggles to adopt AI and life sciences innovations at scale, using personalised medicine as a case study. The first evidence session heard from academics working on genomics, including Professor Sir Mark Caulfield of the 100,000 Genomes Project. Written evidence is open until 20 April 2026.
UK ranks sixth in Public Sector AI Adoption Index 2026
A new index benchmarking public sector AI adoption across ten countries places the UK sixth with a score of 47 out of 100. The index found that 45% of government AI applications are deployed as bolt-on or standalone tools, fewer than one in three public servants use AI to improve workflows, and 63% of civil servants report knowing "a little" or "nothing at all" about AI, even as 60% say they find the tools easy to use.
Nordics
Sweden adopts its first national AI strategy, sets sights on world-leading public administration
Sweden's government has adopted a comprehensive national AI strategy with an explicit goal of being world-leading in the use of AI in public administration. The strategy includes plans for a national AI workshop for public administration, intended to be fully operational by 2030, offering shared infrastructure and a library of reusable AI solutions across government bodies. Sweden is targeting a top-ten global ranking across all AI domains.
Denmark launches AI regulatory sandbox for public and private sector
Denmark's Agency for Digitisation has opened a regulatory sandbox giving public and private sector organisations access to expert guidance on developing and using AI services within the regulatory rules. The move accompanies a reboot of Denmark's national cyber security council, which has expanded its remit to cover AI risk.
Europe
EU Council agrees to delay high-risk AI Act requirements to 2027–2028
As part of the Omnibus VII simplification package, the EU Council adopted a negotiating position that would push back the application of high-risk AI system rules to December 2027 for stand-alone systems and August 2028 for AI embedded in products. The Council also reinstated the obligation to register AI systems in the EU database, added a new prohibition on AI-generated non-consensual intimate content, and delayed the deadline for national AI regulatory sandboxes. The Council's mandate now goes to the European Parliament for negotiation.
Research Spotlight
Building an AI-Ready Public Workforce
The OECD's policy brief examines how AI adoption is changing skill requirements in public administrations and what governments need to do to close capability gaps. Key findings: AI can significantly reduce administrative load (Finland's Kela reports saving 38 years of full-time equivalent case-worker time annually through AI document processing), but most public sector institutions are not systematically investing in workforce capability. The report calls for structured upskilling programmes, governance investment, and an explicit recognition that AI literacy is now a baseline professional skill for public servants.
Why it matters: The report gives public sector leaders a clear frame. The technology gap is secondary to the people gap. Most councils and NHS bodies have AI tools available to them; most do not yet have the training, governance, or managerial ownership needed to use those tools well. This is the thing to fix first.
One to Watch
The GOV.UK Chat team has confirmed it is planning to test agentic AI deployments later in 2026, AI that does not just answer questions but takes actions on behalf of users. This is a significant step beyond what the current chatbot does. No timeline has been given, but it signals that GDS is thinking about AI as an actor in public services, not just a search interface. Public sector leaders who assume AI is "just a chatbot problem" should watch this space closely.
