A recent closed-door roundtable at Parliament, hosted by HCLTech and The Entrepreneurs Network, brought together senior leaders from across the UK Government and the wider public sector to discuss the practical realities of digital transformation and AI.
Opening remarks for the discussion came from Sian Thomas, Chief Data and AI Officer, DBIST, David Caldwell, Deputy Director Service Strategy and Transformation, DEFRA and Paul Montgomery, Vice President - Head of Public Services, Aerospace and Defense, Europe, HCLTech.
The conversation revealed considerable appetite for AI, alongside a clear recognition that enthusiasm for the technology can sometimes move faster than the systems, processes and people needed to support it.
Across public services, the immediate opportunity is substantial. AI is already being explored for administrative tasks, document analysis, drafting, operational processes and service delivery. Yet the suitability of AI varies significantly between use cases. A probabilistic language model may work well for summarizing information or assisting with drafting, while a service determining whether a business qualifies for a specific entitlement requires a much higher degree of certainty.
That distinction places greater emphasis on understanding the problem before selecting the technology.
Start with the foundations
Many of the barriers discussed at the roundtable predate AI.
Government organizations continue to operate with fragmented systems, inconsistent data and processes that remain partly dependent on paper. These conditions make even straightforward digital transformation difficult and weaken the foundations that AI relies on.
Against that backdrop, strong data, infrastructure and business architecture are central to AI readiness. Without them, adding intelligence to existing processes risks producing faster versions of the same underlying problems.
The same principle applies to process design. Automating an inefficient process simply allows inefficiency to operate on a greater scale. Teams need to understand how a service works today, identify where the friction lies and redesign the process before determining where AI adds value.
That work is less visible than launching a new AI service, but it shapes the quality of everything that follows.
AI economics need attention too
The economics of AI are also becoming part of the transformation agenda.
Cloud adoption created new flexibility for organizations, followed by a growing requirement to manage consumption and cost. AI introduces a similar dynamic. The cost of infrastructure is only one component; model selection, token consumption, prompt design and the use of smaller local models all influence the economics of deployment.
“How do you manage the cost of AI? That’s not just about building infrastructure, it’s also about consumption,” said Montgomery.
For public sector organizations, that makes cost an important consideration at the design stage, before AI use expands across departments and use cases.
This also shifts the conversation away from choosing a model purely for technical performance. The most appropriate approach depends on the outcome required, the level of accuracy, the cost of execution and the controls surrounding its use.
Bring people into the transformation
The strongest theme across the discussion was the workforce.
AI adoption changes roles, workflows and career paths. It also raises questions about how expertise develops when routine entry-level tasks are increasingly automated. Organizations still need highly skilled people who understand the underlying systems well enough to intervene when technology fails or produces an unexpected result.
“The pace of change is part of the challenge,” said Montgomery. “There’s a massive skills gap in being able to ask the right questions and make the right decisions as the technology evolves. The speed of change is exacerbating it all.”
That makes skills development most effective when it happens close to the work. Generic online training and centrally published guidance have limited impact when employees are trying to understand how new tools relate to their day-to-day responsibilities. More effective adoption comes from working directly with operational teams, understanding their processes and helping them apply technology to the problems they already face.
The discussion also highlighted a wider opportunity. Many public services already operate with constrained teams, growing caseloads and increasing volumes of information. Productivity gains from AI could give those teams greater capacity to focus on higher-value work and improve the quality of services delivered to citizens.
That requires a clear people strategy alongside the technology strategy.
Build confidence through evidence
Successful adoption ultimately depends on evidence.
AI initiatives need clear outcomes, measurable value and an understanding of where the technology performs well and where human judgment remains essential. Leaders need to know whether a use case improves productivity, service quality or value for money rather than relying on enthusiasm for the technology itself.
That discipline also builds trust. Employees are more likely to adopt AI when they understand how it helps them do their jobs, while citizens need confidence that technology used in public services is reliable and governed appropriately.
The next stage of public sector AI will depend on how well organizations combine technological progress with stronger foundations, better processes, sustainable economics and workforce engagement. Together, these elements create the conditions for AI to improve public services at scale.
