From One School to a Network: What Changes When AI Adoption Has to Scale

The first school has completed its pilot. Staff know the workflow, the person leading it can answer questions quickly and senior leaders can see where support is needed. The results are encouraging enough for the network to consider bringing in more schools.

That is often the moment when a straightforward success becomes a harder leadership decision. The next school has a different timetable, different staff confidence and less access to the people who held the pilot together. Sending over the same tool and training slides will not recreate the support around them.

The practical answer is to slow down before speeding up. Work out what helped the pilot succeed, make those conditions clear for the next school, add schools in small supported groups and improve the plan before the next group begins.

What worked in one school may not travel by itself

A pilot can show that a particular use of AI is worth pursuing in one setting. Moving beyond it means making five things clear before another school starts:

• Shared purpose: Everyone can explain which school problem the use is meant to address and what is outside its scope.
• Clear ownership: Each school has a local lead, the network has a named source of support and people know where to take questions.
• Realistic support: The next school receives the time, training and access it needs, not the unusually close attention available during the pilot.
• Visible boundaries: Staff can see the approved use, the checks they must make and the situations in which they should stop and ask for help.
• A next decision: Leaders agree what they need to learn and when they will decide to continue, change or pause.

The Education Endowment Foundation’s implementation guidance describes implementation as a structured but flexible process shaped by context, people and capacity. It is general school-improvement evidence, not proof for a particular AI rollout. Its value here is practical: the work around an initiative matters as much as the initiative itself.

If the first school has not yet produced evidence strong enough to support a decision, return to a bounded pilot with a clear use case, support plan and measures. Adding schools will not answer questions the original pilot was meant to settle.

Start by finding what made the pilot work

Pilot reviews tend to focus on what people could see: staff took part, outputs were useful, questions were resolved and setup went smoothly. Before expanding, trace those results back to the people, time and support behind them.

These are possibilities to investigate, not claims about why the pilot succeeded. Use notes from the pilot, speak to the people who delivered and used it, and look for the support they relied on even when it was informal.

Write a plan the next school can actually use

The next school should not have to reconstruct the pilot from meeting notes or someone else’s memory. Give it a short, practical plan that covers:

• Purpose: The problem this use of AI is intended to address.
• Everyday use: The staff or student workflow being introduced and what good checking looks like.
• People: Who leads in the school, who supports from the network and where different questions go.
• Preparation: The time, training, access and guidance people need before they begin.
• Progress: What the school will notice, discuss and bring to the next decision.

This plan does not decide everything that should be common across the network and everything that schools should choose locally. That is a separate leadership and governance decision. Its job is simpler: help another school begin well without copying the first school blindly.

The OECD’s implementation framework for effective change in schools brings together a clear strategy, early involvement and room for local adaptation. That balance matters in a network. Schools need enough consistency to learn from one another, and enough freedom to make the approach work in their own setting.

Add schools in small groups, with a clear reason

Bringing in a small group of schools gives the network time to notice where its plan is helpful and where it depends too heavily on the first school’s circumstances. The group should be small enough to support properly. It should also include enough difference to reveal weak assumptions, perhaps in leadership capacity, staff confidence, curriculum context or technical setup.

The sequence will vary by network. What matters is knowing what each group should help you learn.

Do not let a calendar make the decision automatically. At the end of each group, choose one honest outcome: ready to continue, needs a change or more support, or not ready yet. A pause is useful when it prevents the next schools from inheriting a known problem.

Make the next school’s job easier

Scaling should remove repeated work. Schools should not each have to recreate the same briefing, approval route, training sequence and support list.

Build a small shared pack and improve it as you learn:

• A short brief explaining the purpose, the use and its current boundaries.
• A starting conversation covering local ownership, available time and likely obstacles.
• An onboarding plan based on the real workflow, not generic AI training.
• A support guide showing where teaching, technical, data and safeguarding questions go.
• A record of decisions showing important changes, unresolved issues and the reason for local adaptations.
• A few signs of progress that schools can discuss before deciding what happens next.

Reuse should save later schools time. It should not force them to accept an answer that only suited the first school.

England’s multi-academy trust leadership framework offers one system-specific example. It asks trust leaders to understand the contexts of their schools, vary support, identify expertise and capacity, build collaboration and create regular feedback. Networks in other countries work under different structures and rules, but the practical point travels: schools need support that reflects their circumstances, while the network still learns together.

That support should not rest on one enthusiastic colleague. A local champion can help people apply the plan, but named leaders and specialists still need to own the decisions and support behind it. Our guidance on distributing school AI capability explains why enthusiasm should strengthen the support system, not become the whole system.

Show schools what changed because they spoke up

Feedback is useful only when schools can see what happens to it. Keep the process simple:

1. Hear it: Capture the obstacle, local change or missing support while it is still fresh.
2. Act on it: Decide whether to help one school, change the shared guidance or adjust the next group.
3. Close the loop: Tell schools what changed, what did not and why.

IIEP-UNESCO’s work on adaptive implementation in education reform emphasises capacity, participation, feedback and local adaptation guided by shared goals. It looks at education reform rather than AI adoption specifically. Its relevance here is that schools’ experience should shape implementation while change is still under way, not become a report read after the rollout is over.

Look for readiness, not just activity

Licences, logins and training attendance can show whether people have access or took part. They do not show, by themselves, whether a school is ready to continue.

In each school, look for a rounded picture:

• Purpose and ownership: Local leaders can explain why the use matters and who owns the next decisions.
• Staff support: People have enough time, guidance and access to use it as intended.
• Educational fit: Teachers can use it without losing the curriculum purpose or their professional judgement.
• Safe use: Users, information and routes for raising concerns match the school’s approved conditions.
• Available help: The network knows which questions recur and has enough capacity to respond.
• Useful learning: The school can explain what is working, what remains uncertain and what should change.

Use these signs to support a conversation, not to create a universal score. One school may be ready to continue with the same use but not to bring in more staff. Another may need extra time or support before it continues. Seeing that difference is part of responsible scaling, not evidence that the network has failed.

A simple way to decide what happens next

Successful pilots often contain more support than anyone first realises. Leaders make time. A capable person solves problems quickly. The first group gets close attention. Those are not reasons to discount the pilot. They are clues about what the next school may need.

The five moves below are TopSchool’s summary of the guidance in this article. They are not an official framework from the organisations cited above.

TopSchool synthesis: five plain-language moves for network rollout

1. Work out why the pilot succeeded. Look beyond the visible result to the time, people, boundaries and support behind it.
2. Write down what another school needs. Give the next team a clear purpose, named owners, practical preparation and a decision point.
3. Support a small group of schools. Choose a group the network can help properly and use it to find weak assumptions.
4. Improve the plan using what schools tell you. Change the guidance or support, then explain what changed and why.
5. Add more schools only when the network can support them. Continue when schools can begin well, reach help and contribute useful learning.

The aim is not to reproduce the first school. It is to make each new school’s starting point clearer and the network’s support more dependable. That is what allows adoption to grow without asking later schools to succeed on enthusiasm alone.