From ambition to action: what will it really take to make AI work across government?

We’ve been here before. In 2010, Transform co-authored the Martha Lane Fox report, firing the starting pistol on the digitisation of public services. Transform delivered the first exemplar for the Office of the Public Guardian with the digitisation of the Lasting Power of Attorney service. That project helped lay the foundations for Government Digital Services and the standards that have since shaped how government does digital.

AI needs to follow the same path, with central government coordinating communities of practice, setting guidance, developing standards and building tools for everyone to use.

The need for coordinated effort

The UK government has made encouraging moves. The AI Opportunities Action Plan, published in January 2025, sets out a ‘Scan-Pilot-Scale’ framework for public sector AI adoption. The AI Playbook launched in February gives departments accessible guidance. The £2 billion AI commitment in the Spending Review signals that this isn’t just rhetoric.

Without sustained central support, there’s a real danger that individual government bodies end up going on this journey alone. It’s not a simple journey.

Organisations face a daunting set of decisions: how far and how fast to develop with AI; how to align an AI strategy with their organisational strategy; how to identify and prioritise use cases; how to assess their data and technology readiness; how to make technology choices that don’t trap them in a single vendor; how to retain sovereignty over their own systems. There’s also the hardest part – people and cultural change. As with data transformation and digital transformation before it, AI transformation is at least 70 percent about people and culture, and 30 percent about the technology and data.

The more guidance that can come centrally, and ideally the more tools that can be developed and shared centrally, the faster organisations will get there, and at lower total cost to the UK taxpayer.

At Transform, we’re currently working with many of the largest government departments on their AI transformation, and in conversation with many local authorities about what they’re trying to achieve. Our work spans the full journey: creating AI strategies aligned to organisational strategies; assessing data and technology readiness; writing business cases to secure senior leadership, buy-in and funding; building AI tools that protect organisational sovereignty; and new operating models and people change programmes to support when tools go live. Every organisation is in a different place on this journey, and we meet them where they are.

Some examples from our work across Government

This year we developed a data and AI strategies for the National Audit Office, aligned to their five-year organisational plan. They set the direction for how the NAO will use AI to support their own internal efficiency and how they will audit UK government spending on AI over the coming years, which will be significant. Having the right strategy sets the direction and pace for everything to follow.

For years, we’ve been helping a large government department face the ethical and technical challenges of adopting AI as part of their core offering for citizens. We’re working in collaboration with DSIT and i.ai in a complex space where AI can definitely deliver positive outcomes, but with important considerations to manage around safeguarding and standards.

We’re also working with HMRC’s trade team on the end-to-end journey for small and medium-sized businesses looking to import and export. AI has real potential here: helping businesses find accurate tariff and taxation information, and navigating the complexity of trade guidance, the kind of task that can be a barrier to trade.

In Local Government, we are supporting many Councils who are looking to AI not just for efficiency savings, but to genuinely transform the services they deliver for their communities. The LGA’s own research shows AI is already being deployed across adult social care, housing, planning and customer services. The potential is real, so is the risk of fragmentation and duplicated efforts if organisations are left to figure it out for themselves.

Working together

The GDS model gave us something valuable: common standards, shared assessments, and a community of practice that meant good ideas didn’t stay siloed. We need the same for AI – a central function that not only sets standards but actively shares what’s working, and makes it easier for a council in the North East to benefit from a solution pioneered by a Whitehall department.

The technology is constantly improving. The ambition is there. What’s needed now is the connective tissue – the guidance, the shared tools, the knowledge exchange – that turns individual experiments into a genuine, system-wide transformation.

We’ve done this before. We can do it again.