What Should Be Standardised Across a School Network's AI Rollout?
Set common safeguards for AI use across a school network while keeping curriculum, teaching and local implementation decisions close to schools.
What this guide covers5 sections
A network leadership team has two AI requests on its agenda. The first is for teachers to use an approved tool to turn their own notes into ideas for a low-stakes quiz, without entering personal data. The second is for a learner-facing service connected to school accounts, with access to information that could shape support or assessment.
Both requests involve AI, but they should not travel through the same approval process. Treat them alike and the network either slows simple work with unnecessary checks or gives a more sensitive use too little scrutiny. Push every detail to the centre and shared safeguards can also start dictating teaching.
Set the same approval route, tool and information rules, supplier checks, minimum staff expectations, concern route and review record across the network. Let schools decide how an approved use fits curriculum, teaching, daily routines and additional staff support. Require more review when learners, personal data, automated decisions or connected systems are involved.
This is not a universal legal rule. Each network still has to follow the law, safeguarding duties, assessment rules and procurement requirements that apply to its schools. It is a practical way to make responsibility clear without removing professional judgement.
Set the rules that schools should never have to guess
A common rule earns its place when it prevents every school from interpreting the same institutional risk differently. The aim is not to make every school work in the same way. It is to give staff dependable answers where variation would create confusion, repeated work or avoidable risk.
Give every school one route for approving a proposed use
Staff should be able to see who can approve routine use and when a proposal needs specialist advice. Name the leader accountable for the network’s overall position, then show when academic, safeguarding, data protection, IT or procurement colleagues need to be involved.
The route should follow what the proposed use does. A teacher working with an approved tool and no personal data needs a clear boundary and a place to ask for help. A learner-facing service connected to school records needs a more careful decision before anyone begins.
Our guide to school AI governance roles and decision rights explains how to connect one accountable leader with the right advice without sending every proposal to a large committee.
Make the rules for tools, information and access easy to find
Every school needs the same basic answers. Which tools and accounts are approved. Which information must never be entered. Who can access each service. Which systems may connect. What staff should do if a tool changes or something goes wrong.
This needs to be useful in an ordinary school moment. A teacher should not have to interpret a long policy to know whether they can paste a pupil’s work, name or support information into a tool. A school leader should not have to ask three different teams which account staff should use.
UNESCO’s guidance for generative AI in education and research brings privacy, age-appropriate use, human responsibility and educational design into one coherent approach. In England, the Department for Education’s current AI and data protection guidance for schools tells schools to use approved tools and understand how personal data is handled. It also tells them when to involve data protection or IT leads. Networks elsewhere need to apply their own legal requirements, but staff still need one clear set of local rules.
Ask every supplier the same essential questions
Schools should not receive different assurances from the same supplier because they happened to ask different questions. Use one shared check for the product’s purpose, intended users, evidence, data use, security, monitoring, incident reporting, other companies involved, deletion process and important product changes.
The Department for Education’s generative AI product safety standards are mainly written for suppliers in England, although schools can use them when assessing products. They provide a useful example of the detail a provider should be able to explain about purpose, privacy, security, oversight and learner safety.
Completing a shared supplier check does not make a product suitable for every lesson. It removes repeated institutional work. Schools and teachers still decide whether an approved product serves a particular educational purpose.
Give every member of staff a clear starting point
Everyone who may use AI needs enough understanding to recognise an approved use, protect information, check important outputs and know when to stop and ask for help. That is the shared minimum.
It does not require identical training in every school. A network can set the expected judgement and common boundaries while schools choose the examples, sequence and extra support that fit their roles and curriculum. The European Commission’s guidance for educators using AI and data similarly centres critical understanding and decisions made in context.
Our guide to AI literacy for teachers gives a practical starting point based on judgement rather than prompt technique.
Give concerns one clear entrance
A member of staff should not have to decide whether a concern belongs to IT, safeguarding, data protection, assessment or a school leader before they can raise it. Give every school one clear reporting route. The right people can then take responsibility for the response.
State which events require an immediate pause, who needs to know and how the school will receive an answer. Keep a small shared record so the network can notice repeated problems and useful local changes. Counts of logins, prompts or training attendance do not show whether a use remains educationally sound or safe.
Keep curriculum and teaching choices close to schools
Once the common conditions are clear, schools need room to make an approved use work for their learners. Local decisions should include:
- where the use supports the curriculum and where it would remove important thinking or practice;
- how teachers introduce the activity, check outputs and combine it with discussion, feedback or independent work;
- how the use fits timetables, planning and existing school routines;
- which permitted settings, templates or materials suit the learners and subject;
- how quickly the school proceeds and which additional training or coaching staff need; and
- whether not using AI is the better educational choice.
The OECD’s Digital Education Outlook 2026 describes teachers retaining a meaningful role in the design and day-to-day orchestration of AI-supported learning. Shared safeguards should support that judgement, not replace it.
A school should be able to adapt an approved use without seeking central permission for every lesson while five things remain unchanged: the purpose, approved tool, information rules, users and risk conditions. A change to any of those may need another review.
The companion guide on moving from one school to a network covers the separate job of supporting schools in stages and learning from each group. This guide stays with the narrower decision about what must remain common during that rollout.
Match the review to what is at stake
The right amount of checking depends on the people, information and consequences involved. Use the table as practical routing guidance, not as a legal checklist.
| What is changing | Checks the network requires | What the school or teacher still decides |
|---|---|---|
| An approved staff-facing tool is used for low-stakes planning without personal data | Use the approved account, follow the information rule, check important outputs and know where to get help | Whether the tool helps, how to apply the output and whether not using AI is better |
| Learners will use the tool, it supports assessment or it changes an established way of working | Confirm the purpose, age-appropriate access, teacher oversight, academic and safeguarding input and a review point | How the activity fits the curriculum, how it is introduced and what evidence of learning is needed |
| The use involves personal or sensitive data, automated decisions, identity, connected systems or serious safeguarding implications | Involve the right specialists before use, obtain clear supplier evidence, set access controls, agree an incident response and record the decision | Whether the educational case is strong enough to proceed within the approved conditions |
Local law or policy may require additional checks. The practical aim is to stop schools guessing when to ask for help without forcing a simple staff task through the same process as a complex or sensitive use.
Write one agreement people can use
A committee decision is not finished when everyone in the meeting agrees. Record the boundary in a form that school staff can use when the next proposal appears.
Write down five things:
- The shared conditions. State what applies across the network and which people or uses it covers.
- The local choices. Name what school leaders, academic teams and teachers can adapt without another central approval.
- The owners. Identify the accountable network leader, the person responsible in the school and the specialists required when more is at stake.
- The return triggers. List the changes or incidents that bring the use back for review, such as new data, learner access, integration, automation or an important supplier change.
- The review point. Set the date or event for review and the information the committee will examine.
Test the record with one current proposal. A school leader, teacher and IT or data colleague should reach the same route without interpreting the policy differently. If they cannot, make the boundary clearer before the use expands.
Review the boundary, not every lesson
Consistency is useful only while the shared decision still serves schools. Review it when a supplier changes the product, new guidance changes the risk, several schools meet the same obstacle or a local adaptation reveals a better shared safeguard.
That review should not pull every teaching decision back to the centre. If a school remains within the agreed purpose, tool, information, user and risk conditions, local judgement should stand. The network needs to improve the boundary, not supervise every lesson inside it.
Done well, this gives every school dependable protection without making practice uniform. Schools receive the same approval route, information rules, supplier checks, starting expectations and route for help. They keep responsibility for turning an approved use into sound teaching and workable local practice.
Standardise what schools should never have to guess. Keep educational choices with the people who understand the learners, curriculum and daily work.
