AI in the Classroom: Safe, Useful and Fair
Treat AI as a draft partner and a tutor—always under teacher judgement and with clear guardrails.
Challenge
Schools wanted the benefits of AI tools without compromising safeguarding, assessment integrity or equity.
Result
A practical AI policy, teacher prompts bank and pupil usage norms enabled safe, effective classroom use.
Outcome
Clear gains in planning time and explanation quality, with integrity protected and digital literacy improved.
Innovation
AI playbook with roles and rules, prompt libraries tied to curriculum, and integrity checks embedded in assessment.
Brief overview
The question is not whether to use AI, but how to use it well. We made roles explicit: AI as draft partner for staff, as a tutor for pupils, and as a tool that never replaces teacher judgement.
Mechanisms that move practice
Staff used prompts to generate drafts for models and retrieval; pupils used AI tutors to rehearse explanations. Assessments used mixed formats and oral checks to protect integrity.
Human moments that matter
A teacher saved 30 minutes on a model draft, then improved it; a pupil practised an explanation with instant feedback before presenting.
Keeping workload net zero
Shared prompt banks and model formats reduced reinvention. Integrity checks were built into existing routines rather than added as extra tasks.
Evidence and alignment
We tracked planning time saved, clarity of models and the rate of integrity flags. Pupil voice captured confidence with AI literacy.
Impact
Teachers reported better explanations; pupils practised more; integrity concerns were handled calmly with clear procedures.
Lessons for leaders and investors
- Publish a short AI policy with roles and rules.
- Tie prompts to live curriculum products.
- Build integrity checks into assessment design.
- Teach digital literacy explicitly.
Full Article
AI has moved from novelty to tool. The challenge for schools is to capture value without losing integrity or overwhelming staff. Our approach made roles explicit and kept teacher judgement at the centre.
We wrote a short AI playbook that fit on two pages. It set roles and rules: staff could use AI to draft models, questions and letters; pupils could use approved tutors for rehearsal and explanation; teachers remained the editors of everything that entered classrooms. The playbook also set out data protection expectations and routes for concerns.
Prompt banks lived next to curriculum maps. For each unit, teachers had a short list of tested prompts that produced model paragraphs, worked examples or retrieval questions. Because prompts were tied to real products, outputs were easier to edit and quality rose quickly.
Pupils used AI tutors in bounded ways. In the Angola hybrid model, mornings provided adaptive practice based on recent teaching. The same principle applied here. Pupils rehearsed explanations, checked understanding and received hints, not answers. Sessions were logged so teachers could see themes and intervene humanely.
Assessment design carried the integrity burden. We used mixed formats—seen questions that required application to new material, practicals, viva‑style checks and supervised writing. Where AI could help a pupil prepare, that was a feature, not a bug; where it could short‑circuit the product, the assessment changed.
We trained teachers to read AI. They learned to spot the plastic quality of some outputs and to insist on artefacts of thinking: plans, drafts and oral rehearsals. Colleagues used AI to produce first passes, then rewrote for clarity and accuracy. The standard was not novelty but whether the result helped pupils learn.
Human moments persuaded sceptics. A history teacher who distrusted AI found that a prompt produced a clean, editable model paragraph in minutes. A pupil who feared presentations ... Back the AI playbook, prompt banks and assessment design time. Expect benefits in planning time and explanation clarity, with integrity protected. Because the approach is behavioural and artefact‑based, it scales across phases and contexts. Two‑page AI playbook; unit‑linked prompt banks; bounded pupil tutor use; integrity‑conscious assessment design. Teachers felt in control; pupils felt supported; families trusted the guardrails. Reuse prompts; embed checks in existing routines; avoid parallel systems. Aligned with DfE data protection advice, EEF guidance on feedback and explicit instruction, and inspection expectations for assessment integrity.What this means for school leaders and investors
Full narrative expansion
What changed in practice
Human moments that built culture
Results we saw
How we kept workload net‑zero
Evidence and UK alignment
External Links