UK Government AI Forum Urges Service Redesign and Stronger Procurement Safeguards

At Think AI for Government 2026, public-sector and industry speakers called on UK departments to treat artificial intelligence as an opportunity to reconsider how services are delivered—not merely a tool for speeding up existing processes. The discussion included HM Revenue & Customs (HMRC), the Ministry of Housing, Communities and Local Government (MHCLG), ElevenLabs, and the Ada Lovelace Institute. Although the event produced no specific solicitation or contract award, its themes offer useful indicators for companies preparing to compete for future government AI work: proposals may need to show how technology improves outcomes for citizens, supports service redesign, and fits the department’s operational and governance requirements.

A central point was that automation can preserve problems embedded in existing services. Nuala Polo of the Ada Lovelace Institute said, “We’ve done a lot of research on how we want to envision public services rather than automating the services that we already have, which might not work for people, and making those things faster.” Giuseppe Sollazzo, deputy director in HMRC’s chief AI office, similarly emphasized starting with users and their needs, invoking the Government Digital Service’s user-centered approach. He cautioned that transferring a process into a digital format can impose unnecessary restrictions, citing online forms, and argued that AI can help teams brainstorm entirely different service approaches. For bidders, this suggests that a credible solution should connect its technical functions to a defined user need and explain whether it enables a better service model—not only a more efficient version of the current one.

The panel also highlighted digital sovereignty, public trust, and the risks of dependence on AI models developed by overseas technology companies. Polo said that government dependence on such models could put the public sector “in a really delicate place,” and questioned whether every application requires the most advanced models when smaller alternatives might be appropriate. This does not establish a new procurement rule or indicate that foreign suppliers are excluded. It does, however, signal that vendors should be prepared to explain where data is processed and stored, what protections apply, how dependencies are managed, and whether a proposed model is proportionate to the use case. The source also identifies data sovereignty as a significant consideration for ElevenLabs’ UK&I lead, Elias David, reinforcing its relevance as a supplier-facing issue.

Operational readiness and adaptability were also prominent. Chris Clark, MHCLG’s deputy director of advanced analytics and AI, described work with local authorities to understand how AI might address service challenges, alongside the cultural and organizational change needed to sustain adoption. He said organizations that are most adaptable are likely to adapt best. For procurement teams and contractors, that points to implementation plans, staff capability, change management, and continuing system oversight as considerations alongside model performance. The event’s emphasis on procurement expertise and workforce readiness likewise suggests that departments will need people able to assess AI systems and manage them after deployment. Suppliers can strengthen future bids by presenting evidence of testing, plans for monitoring ongoing performance, and approaches that can accommodate changing requirements. These are market signals from a discussion, not announced evaluation criteria or a live opportunity.

• No solicitation or award was announced at Think AI for Government 2026; the discussion is an indicator of possible future buyer priorities, not an active procurement.
• Buyers and contractors should frame AI around user-centered service redesign, rather than automation or digitization alone.
• HMRC’s Giuseppe Sollazzo emphasized asking what users need and noted that existing processes, including online forms, can impose unnecessary restrictions.
• The Ada Lovelace Institute’s Nuala Polo called for reimagining public services and cautioned against making unsuitable processes faster without redesigning them.
• Digital sovereignty and reliance on overseas AI models were raised as concerns; suppliers should be ready to address data handling, dependencies, and public trust.
• Polo questioned whether government applications always need the most advanced models, creating a potential case for right-sized solutions where suitable.
• MHCLG’s Chris Clark highlighted organizational adaptability and work with local authorities on AI-enabled responses to service challenges.
• Future offerings may need to substantiate testing evidence, data protections, and ongoing performance, while explaining how systems can adapt as requirements evolve.
• Procurement expertise, workforce readiness, implementation planning, and staff capability may matter alongside the technology itself.