The UK Government’s AI Opportunities Action Plan sets a broad direction for accelerating AI adoption, strengthening enabling capabilities and using AI to improve public services. Early information is already being published. But the harder task is operational: turning national ambition into redesigned services that are measurable, reusable and trusted.
This is the missing middle of the AI debate: the space between national ambition and operational reality. National plans create momentum, but they do not, on their own, change how a service is organised, governed or experienced.
The current AI narrative is still too technology led. It asks whether the public sector can adopt AI safely, but not often enough whether AI should force a redesign of how services are organised, funded, staffed and measured.
Why AI needs an operating model, not just a use case
Public services are under structural pressure across health, social care, local government, policing, and justice. Demand is rising, needs are more complex, and many organisations are still working around legacy systems, fragmented data and outdated processes. If AI is simply bolted onto those processes, it may make parts of the system faster, but it will not change the economics of delivery.
In this context, operating model means more than a process map. It includes roles, decisions, evidence, data, controls, and reuse mechanisms. AI creates lasting value when these elements change together.
Reinvention starts with redesigning whole service pathways, not automating individual steps. AI-enabled triage, for example, could help people understand eligibility earlier, route complex cases to specialists, reduce avoidable contact and give teams better evidence about demand. The aim is a simpler, more proactive and more responsive service.
It also means building shared capabilities across authorities and departments. Many public services face common problems such as evidence assessment, application processing, demand forecasting and case prioritisation. Solving these repeatedly as separate projects creates duplicated cost and uneven maturity. AI creates the opportunity to turn repeatable capabilities into reusable public service components.
The hard part starts after the pilot
For public service leaders, the test is uncomfortable but simple: if an AI pilot succeeds technically, what would you be prepared to change operationally so that it can matter at scale?
The AI Opportunities Action Plan rightly points towards the need to move beyond isolated experimentation and create the conditions for scaled adoption. But scale will not come from pilots alone. It requires strong data foundations, clear accountability, user-centred design, operational ownership, assurance and evidence that services improve outcomes.
Police AI points to what this shift could look like: a more coordinated route to identify, test and scale AI tools across policing, with greater emphasis on shared capability, assurance, operational use and public trust.
The next phase should be judged differently. Not by the number of AI tools trialled, but by the number of services redesigned. Not by novelty, but by evidence of reduced cost, improved quality, faster access, better decisions and greater citizen trust.
The principles that turn AI into public value
- Redesign the service pathway – Start with the outcome, not the technology.
- Strengthen the data foundation – Improve data quality, interoperability, governance and access.
- Reuse shared AI capabilities – Avoid solving common problems repeatedly in separate silos.
- Build trust into the operating model – Make decisions transparent, auditable and accountable.
- Measure public value – Track outcomes, cost avoidance, service quality and citizen experience.
This moves the conversation from AI use cases to AI-enabled change. Leaders need to ask which services should be redesigned, which capabilities should be shared, which controls are needed to sustain trust, and which outcomes prove that the model is working.
The organisations that lead will be those that publish reproducible, metric-driven examples of AI-enabled service redesign. They will show not only that AI can work, but that it can make public services more accessible, sustainable and effective.
The real prize is not AI adoption. It is public service reinvention. Creating operating models that make better outcomes affordable, trusted and scalable. If the sector wants AI to change public services, it must be willing to change the services around AI.
If you have a question for Dave or the Triad team, please get in touch.
