We've experimented, now we need to scale
Moving beyond proof-of-concept through purpose, people, and practical action
You're in good company
If you have run successful pilots but cannot turn them into enterprise impact, you are in the company of most public sector organisations one year on. The Action Plan: One Year On is candid: tools like Minute and Extract were proven in pilot last year, and the priority through 2026 is national scale-up. Scaling is a different discipline to piloting, and most of us are learning it at the same time.
38 of 50 AI Action Plan commitments delivered in year one
DSIT, Jan 2026
93% of teachers across International Schools Partnership completed AI training as part of a multi-year, 25-country rollout
Version 1 + ISP, AWS-based
6th of 10 UK ranking on the Public Sector AI Adoption Index 2026
47/100
“The challenge now is how to adapt organisational design, capabilities and culture to keep pace. AI optimises brilliantly, but within constraints that you and your teams set.”
Version 1 ‘From Playbook to Progress: Delivering AI in the Public Sector’
Understanding where you are
Pilots stall when organisations focus on the technology and not the system around it. The government's own evidence bears this out.
The AI Coding Assistant trial run by GDS across more than 50 public sector organisations found developers saved an average of 56 minutes per working day. The productivity gain was real and measurable. And yet 2,500 licences were distributed and only 1,900 assigned. The technology worked. The system around it was not ready.
That is why GDS has moved from trial to sustained programme. The AI Engineering Lab, launched in early 2026, is rolling out licences across all public sector departments alongside training, integration support and shared evaluation because a one-off experiment does not create lasting capability.
The teams we have helped scale successfully share one habit: they keep coming back to the business problem and the user. What problem are we actually solving? Whose work changes? What does a citizen experience differently when this lands? Without that clarity, scaling is deployment at greater volume, and deployment without purpose is what creates the proof-of-concept graveyard most departments are now trying to escape.
Three things that will move this forward this quarter:
1. Stand up a cross-function scale forum within 30 days. Named leads from technical, security, data, legal, operations and end-user representation. Chaired by the executive sponsor, or if one does not exist yet, this is the moment to appoint one.
2. Pick two or three end-to-end use cases tied to a published departmental objective. Stop the rest. Concentration beats coverage when scaling. Pruning is part of the discipline. For each use case you retain, define upfront what success looks like and instrument the data to measure it.
3. Commission a REACH readiness baseline. A short diagnostic across leaders, managers and teams to show where sponsorship visibility and trust gaps actually sit.
REACH
At this stage: Recognise and Engage
Recognise / Rally
Make sure leaders, delivery teams, security, data and end-users all understand why this matters now. Without that, every new conversation starts from scratch.
Engage / Enable
Surface what's making people hesitant and remove it. Rarely about training; usually about workload, role clarity, sponsorship visibility and trust.
What success looks like
Strategic alignment
A clear AI strategy connected to departmental objectives and citizen outcomes. Worked back from the user need, not forward from the tool.
Sponsorship that spans the business
Visible commitment from across the leadership team — executive, technical, security, data, legal and end-users — not a single champion.
A focused use-case portfolio
Two or three end-to-end workflows where AI demonstrably improves the citizen or staff experience, not a scatter of disconnected pilots.
Cultural Readiness
Teams who understand why change is happening and want to participate. Trust is built before scale, not during it.
Practical actions for your teams
- Start with the problem and the user. If you're stuck, go back to basics. What are we solving, for whom, what changes when we get it right?
- Co-create with the people who will use it. Develop use cases with the staff and citizens whose work or experience changes. They hold knowledge no requirements document captures.
- Build a coalition, not a champion. Advocates across delivery, security, data, legal, operations and end-users. AI doesn't scale through one sponsor.
- Use the AI Knowledge Hub. Real implementation lessons from across government on ai.gov.uk. Build on what others have learned.Join the cross government AI community here.
- Communicate relentlessly. Share early wins, share what didn't work, explain the why repeatedly. Silence breeds rumour.
VERSION 1 IN ACTION
International Schools Partnership: scaling AI across 25 countries
Version 1 worked with International Schools Partnership (ISP) on a multi-year programme to put AI tools into the hands of teachers across the global network. Four areas: lesson planning, marking, tutoring, and measuring learning impact. The platform is built on AWS and designed around teacher workflows, not bolted on top. 93% of ISP teachers completed AI training as part of the rollout, with capability built alongside the technology rather than after it.
The story is in how ISP scaled — measured rollout, training in step with deployment, governance from day one, and a clear connection between AI and the educational outcomes it was meant to support.
