An AI Rollout Plan for Multi-Academy Trusts

Trusts are well placed to get AI right: they can make decisions once, negotiate licences at scale and share what works. They can also get it badly wrong, with every school doing its own thing. Here is a phased plan that works.

Phase 1: Set the foundations (half a term)

• Name an owner at trust level, with a small cross-school group. See setting up an AI steering group in a MAT.
• Write a trust AI policy with space for school-level detail. See how to write an AI policy.
• Agree a trust-wide approved tools list and a process for adding to it. See how IT controls AI tools.
• Complete central DPIAs for trust-wide tools, so schools do not each repeat them.
• Brief the trust board on risks, opportunities and the plan.

Phase 2: Choose the platform (half a term)

Most trusts already run Microsoft 365 or Google Workspace across their schools. Starting with Copilot or Gemini on trust accounts gives staff protected AI at no extra licence cost. Add specialist tools only where there is a clear gap — for example, a planning tool for primary or a leadership tool for SLT. Trust-level buying power helps here: see how much AI costs for schools.

Phase 3: Pilot (one term)

1. Pick two or three schools with different contexts — primary, secondary, a school with a strong SEND profile.
2. Target a small number of high-workload tasks: reports, letters, resource adaptation, planning.
3. Measure time saved and staff confidence before and after.
4. Collect prompts and examples that work into a trust library.

Phase 4: Train at scale (one to two terms)

• Train champions in each school, then run whole-staff sessions using the pilot schools' real examples. See what to include in an AI INSET day.
• Include safeguarding and data protection in every session.
• Give specific training to office teams, SENCOs and leaders — the roles with the biggest admin load.

Phase 5: Measure and review (ongoing)

• Time saved on target tasks.
• Staff usage and confidence.
• Data protection incidents — ideally none.
• Annual filtering and monitoring reviews covering AI. See KCSIE filtering and monitoring.

Mistakes trusts make

• Buying licences before training.
• Letting each school choose different tools.
• Treating AI as an IT project rather than a teaching, safeguarding and workload project.

What the trust board needs to see

• The strategy in one page: purpose, platforms, timeline.
• The risk register entry for AI, including data protection and safeguarding.
• Budget: licences, training, staff time.
• Measures of success and when they will be reported.

Working with school leaders

Central decisions work best when heads help shape them. Involve a head from each phase in the steering group, let schools choose which workload tasks to target first, and share wins across the trust quickly — a head seeing another school save hours on reports is more persuasive than any trust memo.

A 12-month timeline at a glance

1. Months 1–2: governance, policy, approved list, DPIAs.
2. Months 3–4: platform decisions and configuration.
3. Months 4–7: pilot schools.
4. Months 7–11: training at scale.
5. Month 12: review, report to board, plan year two.