How to Choose AI Tools for Schools: 7 Questions to Ask
A practical guide for school leaders to evaluate AI tools for teaching, learning, safeguarding and school-wide use.
By Daniel Pearson, Head of AI and Digital Innovation at Hurtwood House and Director of Campus Mind. Published 19 October 2026.
Choosing AI tools for schools starts with a practical question: what do you need the technology to help people do?
The next step is to test that purpose against educational value, school control, data protection, safeguarding and the full cost of using the service. An impressive demonstration is useful, but the decision should rest on how the proposed setup would work for your staff and students.
Drawing on my experience introducing AI at Hurtwood House, these seven questions offer a framework for comparing tools and asking suppliers for meaningful evidence. Teachers can use them to assess a classroom resource; school leaders can use them to evaluate a wider platform.
The seven questions at a glance
- Does it give staff and students the right kind of help?
- Does it help students think for themselves?
- Can your school shape how the AI is used?
- Are the data protections clear and documented?
- What do safeguarding and oversight look like in practice?
- What can you learn from its use in a real school?
- What will it cost to introduce, manage and maintain?
Use these questions during your next AI supplier demonstration or share them with your school's leadership team.
1. Does it give staff and students the right kind of help?
Write down two or three tasks before comparing products. Preparing a lesson resource, finding an approved school policy and helping a student practise a subject each require different kinds of assistance.
For a teacher, a useful result might be a draft quiz that is accurate, appropriate for the class and quick to check. For a student, it might be a question that helps them recognise a misunderstanding and attempt the next step themselves.
Choose a familiar task and test it with the permissions your users would receive. Check accuracy, curriculum fit, accessibility and the amount of editing needed. Include colleagues with different levels of technical confidence.
A focused staff tool may be sufficient for one task. A school-wide platform needs to demonstrate how it serves different groups without giving everyone the same access or behaviour.
Check the service's age conditions. DfE guidance on AI in education calls for appropriate safeguards for pupil use and professional judgement when checking outputs.
Ask in the demonstration: Can we try our own task using the setup our staff or students would actually receive?
2. Does it help students think for themselves?
A polished answer is not enough to judge a learning tool. Examine what the student contributes during the interaction.
Try a familiar homework-style question, then ask the tool to provide the finished answer. Does it immediately complete the work, or ask what the student has tried and offer a useful hint? After the interaction, could the student explain the method or tackle a similar question independently?
The DfE's generative AI product safety standards include expectations that learner-facing products provide help progressively and prompt learners to contribute, rather than supplying complete solutions by default.
This is central to Campus Mind's approach: AI that makes students think. Hurtwood's published account of Hurtwood Mind describes a system designed to ask about students' understanding and challenge them to explain their reasoning. It illustrates the intended approach, rather than proving that every interaction will succeed.
Ask in the demonstration: Show us how the tool responds when a student tries to bypass the thinking.
3. Can your school shape how the AI is used?
Ask which settings your school controls, which the supplier manages and which cannot be changed. Check how users are added and removed, how administrator access is protected and whether different groups can receive different tools and permissions.
Then consider the information an assistant uses. For a school policy assistant, provide an approved sample document and ask a question it answers. Next, ask something the document does not cover. Does the tool identify the gap, or invent a plausible policy?
Check whether staff can find and verify the source. Agree who maintains the documents and what happens when a policy changes.
Campus Mind can use approved school documents as context for custom agents, with different roles and school-specific guidance. Our guide to introducing AI in schools explains the preparation behind that approach.
Ask in the demonstration: Which decisions can our school make about access, resources and AI behaviour, and who maintains those settings?
4. Are the data protections clear and documented?
Ask the supplier to explain the journey of a prompt, an uploaded file and the resulting answer. You need to understand which organisations process the information, where that happens, whether content is used for model training, who can access it and how deletion works.
Treat training, storage and retention as separate questions. A commitment not to train models on school data does not, by itself, explain how long conversations or backups are retained.
Bring your data protection officer into the review. Where a supplier processes personal data on the school's behalf, the ICO explains the requirement for a written controller-processor contract. A data protection impact assessment is required before processing likely to create a high risk to people's rights and freedoms.
Campus Mind does not submit school content for model training or fine-tuning. Provider handling depends on the selected service, features, terms and settings. Our GDPR and data protection information explains the approach and the arrangements to discuss for your school.
Ask before approval: Can our DPO review the processing arrangements, contractual terms and retention settings for this specific setup?
5. What do safeguarding and oversight look like in practice?
Ask how the tool handles unsuitable requests and fictional safeguarding disclosures. Check typed messages, uploads and any other enabled features. Look at both the response the user receives and what happens afterwards.
The DfE's product safety standards set expectations for learner-facing products covering filtering, recording interactions and alerting appropriate supervisors to harmful content or possible safeguarding issues. Ask for evidence against the relevant standards, including any gaps.
Distinguish access to chat history, automated alerts and an actual staff response process. They are different functions. Establish who receives notifications, how quickly, who reviews them and what happens outside school hours.
Campus Mind provides school-authorised access to chat logs where enabled; it does not provide an automatic safeguarding-alert service. That distinction is explained in our data protection information. Chat review alone does not establish that alerting expectations are met, so assess the complete proposed setup before pupil access.
Ask in the demonstration: When this fictional concern is raised, who would know, when would they know and what action would follow?
6. What can you learn from its use in a real school?
Look for a case study that explains what the school wanted to achieve, what was introduced and how staff evaluated it. Ask which features are available in your proposed package and which were built for that particular school.
A useful example should help you judge suitability, not just show that somebody else has bought the product. Separate accounts of implementation from evidence of improved learning or reduced workload. For any claimed result, ask how it was measured and over what period.
Our Campus Mind implementation case study describes a staff-first introduction, followed by wider student use. Practical applications included policy and procedure support, lesson preparation and administrative tasks. Those examples can help you identify a sensible trial, but they are not a promise of identical results in another school.
Ask the supplier: Can you show us a comparable school use case and explain what it demonstrates, including its limitations?
7. What will it cost to introduce, manage and maintain?
Compare the full cost of the proposed service. Request a breakdown of licences, setup, AI usage charges, training, support and any additional features. Establish what is fixed, what varies with use and what happens when a budget or usage limit is reached.
Include the work inside your school. Who adds accounts, maintains documents, reviews interactions and checks outputs? What can staff manage themselves, and when will they need the supplier's help?
Ask about renewal terms, notice periods and service exit, including exporting school content and closing accounts. A successful trial should make the ongoing commitment clearer.
Start with a small staff group, agreed data-handling arrangements and sample information. Compare the time spent on the whole task, including checking and corrections. Judge usefulness and quality alongside cost, rather than counting how quickly the AI generates text.
Ask before committing: What would a realistic first year cost, and what would our staff need to do to keep it working well?
What about Ofsted approval?
Ofsted says it does not directly evaluate individual AI tools. It may consider how a school's use of AI affects children's and learners' experiences and outcomes. A product being used in an inspected school is therefore not evidence of Ofsted approval.
For any certification claim, check who carried out the assessment, what it covers and whether it applies to the service you are considering.
Turn the answers into a school decision
For each shortlisted tool, record the evidence, any unanswered questions and the person responsible for resolving them. Decide whether it is suitable for the intended use, suitable only after specified actions, or not suitable for that use.
A strong choice should have a clear educational or operational purpose, appropriate protections and a manageable plan for the people who will use and oversee it.
The question is not simply whether the AI can produce an impressive answer. It is whether your school can use it purposefully, responsibly and sustainably.
Explore Campus Mind for your school
Built by experienced educators, Campus Mind brings together school-specific assistants, different staff and student experiences, and practical implementation support. Our starting point is what your school wants to achieve, with student thinking and school control at the centre.
Bring these seven questions to a conversation with us. We can explore your priorities, demonstrate the relevant experience and discuss the configuration, data arrangements and support your school would need.
Request a Campus Mind demonstration and discuss a trial.
The DfE and Ofsted guidance referenced here applies to England. Schools elsewhere in the UK should also consult their national education guidance.