Who this is for
This handbook is for two kinds of reader:
- Classroom teachers who want to reclaim their evenings without compromising the quality of their teaching.
- Head teachers and school leaders who are being asked for an AI strategy, an AI policy, and an implementation plan — often before they've had time to build confidence in the subject themselves.
You don't need to read it cover to cover. The teacher section and the leadership section are parallel and independent. The shared foundation at the front and the shared rules at the back apply to everyone.
Pick your path
Or read on for the shared foundation: the evidence base, then the rules that apply to everyone.
What the evidence says
Before the prompts, the short version of what the research tells us. The conversation about AI in schools has been shaped more by hype and panic than by evidence — these three findings are the ones to anchor on.
Click each heading to expand. Collapsed by default so you can move straight to the practical sections.
For Teachers
The mindset shift
AI is not a magic lesson-writer. It's a competent junior colleague who's read everything, knows nothing about your specific class, and needs clear instructions.
That means:
- You are still the expert. AI produces a draft. You apply judgement.
- Specificity wins. "Write me a lesson plan" will get you mush. "Write me a 45-minute lesson plan for Year 8 on photosynthesis, including a starter activity, three differentiated tasks, and an exit ticket — following AQA specification" will get you something usable.
- Iterate, don't accept. The first output is rarely the final output. Ask it to refine, soften, shorten, extend. It won't get tired.
Use case 1 — Differentiation at speed
The single highest-impact use of AI for classroom teachers. Differentiation has always been the right thing to do and the first thing to be cut when you're running out of evenings.
Reading-level differentiation
I'm teaching this passage to a mixed-ability Year 8 class. Please rewrite it at three reading levels: foundation (ages 9–10), core (ages 12–13), and stretch (ages 14+). Keep the key vocabulary consistent across all three. Here's the passage: [paste text]
SEND-specific adaptation
Rewrite this worksheet for a Year 8 pupil with dyslexia. Use shorter sentences, break up dense paragraphs, and add visual cues where helpful. Preserve the learning objective.
Visual sequencing
Turn these learning objectives into a visual step-by-step sequence suitable for a pupil who processes better with pictures than with text. Describe each frame so I can sketch or source images.
EAL support
Produce a glossary of the key technical vocabulary from this text, with definitions written at a B1 English level. Include an example sentence for each term.
Worked example
Input: A Year 9 History lesson on the causes of the First World War, currently pitched at the middle of the class.
Prompt: "Take this lesson and produce three versions of the starter activity — one for a pupil who struggles with reading density, one for the main cohort, and one for a pupil who needs stretching toward GCSE-level analytical thinking. Keep the learning objective identical across all three."
What you get back: Three starter activities, roughly two minutes of AI time. What used to take an hour of your evening now takes five minutes of review.
Use case 2 — The difficult parent email
Every teacher has one. It's been sitting in drafts for three days because the stakes are too high to get the tone wrong.
I need to email a parent about a difficult situation. Here are the facts: — [Brief, factual bullet points — no names, no pupil identifiers] — [What's happened] — [What I'd like the parent to do] Please draft this email three ways: 1. Gentle — assumes good faith, softens the concern 2. Firm — clear expectations, direct but professional 3. Balanced — somewhere between the two All three should be warm, professional, and no longer than 150 words.
Why it works
You get three angles on the same message. Pick the one closest to what you want to say, then edit. The hardest part of the email — the blank page and the first draft — is gone. Your judgement stays exactly where it belongs: on the final version.
Use case 3 — Report writing and end-of-term comments
Report writing season is a known pain point. AI doesn't replace your professional judgement on a child — it removes the friction of turning a bullet-pointed assessment into fluent prose.
I'm writing end-of-year reports for a Year 7 class. Here are my bullet-point notes for one pupil: — Engagement: high, often first to contribute — Written work: strong ideas, sometimes rushed in execution — Collaboration: supportive, helps peers — Target: more careful checking of written work before submission Please produce a 120-word report comment in a warm, professional tone appropriate for parents. Avoid generic phrases like "a pleasure to teach."
Variants
- "Rewrite this in a more formal tone."
- "This is the third pupil with a similar profile — vary the phrasing so all three reports don't sound the same."
- "Add a sentence highlighting the specific target action for the next term."
Use case 4 — Subject-matter comprehension at pace
For ECTs, cover teachers, and anyone teaching out of specialism. AI produces a briefing-quality summary of a topic in minutes.
I need to teach a Year 10 lesson on [topic] but it's not my subject specialism. Please give me: 1. A 200-word plain-English summary of the key concepts 2. The three most common misconceptions pupils hold about this topic 3. Two worked examples at GCSE level 4. Five good questions I could ask to check understanding Follow the [AQA / Edexcel / OCR] specification where relevant.
Use case 5 — Starters, quizzes, and extension tasks
The grunt work of lesson preparation — the material around the main activity — is what AI does best.
For a Year 9 English lesson on [text], please produce: — A 5-minute starter activity that gets pupils thinking about [theme] — A 10-question quick-fire quiz covering the main plot points of the first three chapters — Two extension tasks for pupils who finish early, pitched at GCSE analytical level Format as something I can print directly.
Use case 6 — Marking rubrics and feedback scaffolds
Useful for consistency across a class and across a department.
I'm marking Year 11 essays on [topic]. Please produce: 1. A rubric with four bands (F–C, C–B, B–A, A–A*) covering argument, evidence, analysis, and written expression 2. Three feedback comments for each band that I can adapt — specific, actionable, and written in a supportive tone Follow the AQA GCSE English Literature assessment criteria.
For Head Teachers and School Leaders
The mindset shift
AI is not a replacement for leadership judgement. It's an analyst who can read fast, summarise accurately, and draft competently — then hand the decision back to you.
For school leaders, the highest-value applications are not in the classroom. They're in the policy, compliance, and communication layer where you're currently spending time that should be spent on strategy and on people.
Use case 1 — Policy gap analysis
The single biggest time-saver for most school leaders. You have existing policies. The regulatory frameworks keep changing. AI bridges the gap.
I'm going to paste our school's [e.g. Safeguarding / AI / Online Safety / Behaviour] policy. Please compare it against [the latest KCSIE / the new Ofsted framework / the DfE SEND white paper] and identify: 1. Sections that are out of date or missing required elements 2. Language that should be updated 3. New areas the current policy doesn't cover but should 4. Specific paragraphs I should review first Present as a prioritised action list.
Use case 2 — First-draft compliance documents
Risk assessments, DPIAs, safeguarding updates, EHCP reviews. The blank page problem is real and it eats leadership time.
I need to produce a first-draft Data Protection Impact Assessment for [describe activity]. Please structure it using the ICO's recommended template, and include: — Nature, scope, context, and purposes of processing — Necessity and proportionality assessment — Identified risks to data subjects — Mitigation measures Where you need information from me that you don't have, flag it clearly with [BRACKETS] so I can fill in the gaps.
Why the brackets work
The "flag what's missing" instruction is the key move. AI won't hallucinate your school's specific circumstances — it will explicitly mark where your input is required. That's exactly the division of labour you want.
Examples of good first-draft candidates
- Risk assessments for trips, events, or new activities
- DPIAs for new software or data processes
- Initial Ofsted self-evaluation sections
- Governor report drafts
- Parental consent forms
- Safeguarding updates after KCSIE revisions
Use case 3 — Meeting and inspection-report digestion
You have a 60-page inspection report, a 40-minute meeting transcript, or a pile of SLT minutes from the term. You need the signal, not the noise.
I'm going to paste [document type]. Please produce: 1. A one-paragraph executive summary 2. Key decisions made 3. Actions with owners (if identified) 4. Risks or concerns flagged 5. Any questions or issues left unresolved If the source material is ambiguous on any of these, say so rather than guessing.
Use case 4 — Communications at scale
Parent newsletters, staff briefings, governor reports. The weekly rhythm stuff that eats time that should go on strategy.
I need to write this week's parent newsletter. Here are the key items to cover: — [Bullet list of items] Please draft a 400-word newsletter with: — A warm opening paragraph — Each item as a short clear section with a subheading — A closing line that reinforces our community values Tone: friendly but professional. Audience: parents of pupils aged 11–18.
Customising tone
You can teach AI your voice over time. Paste in three past newsletters you were happy with and ask: "Analyse the tone and structure. From now on, match this voice when I ask you to draft school communications."
Use case 5 — Data pattern-spotting
AI can look at structured data — attendance, behaviour, attainment — and surface patterns that are hard to see at a glance.
I'm going to paste anonymised attendance data for Year 10 across the last two terms. Please identify: 1. Pupils with concerning patterns (consistent decline, unusual absence clusters) 2. Any day-of-week or period-of-day patterns 3. Cohort-level observations 4. Questions I should ask our attendance lead The data has been stripped of all names. I'll cross-reference pupil identifiers myself.
Use case 6 — SENCO and EHCP documentation support
This one sits at the intersection of leadership and SEND provision. It's also one of the most time-consuming areas of school administration.
I'm reviewing an EHCP for a Year 8 pupil. Here are the current provisions and the SENCO's observation notes from this term: [paste anonymised content] Please: 1. Identify any provisions that no longer match the pupil's observed needs 2. Flag any gaps between the statutory entitlement and the documented provision 3. Suggest specific language for the annual review based on the observations 4. List any areas where external professional input would strengthen the case
Why this matters
The EHCP system is under unprecedented strain. AI won't fix that at the system level — but it can make the hundreds of hours of EHCP work inside your school significantly more efficient, and more consistent.
The Rules (For Everyone)
These apply to every use case above, without exception.
Rule 1 — Never input personal data
Hard rule: No pupil names. No staff personal details. No identifiable information of any kind.
Use codes, initials, year groups, and role descriptors:
- ❌ "Jamie Smith, Year 9, has been struggling..."
- ✅ "A Year 9 pupil in my tutor group has been struggling..."
This is a DfE hard line, a GDPR hard line, and a safeguarding hard line. No use case is worth breaching it.
Rule 2 — Always check the output
AI hallucinates. It will invent:
- Dates
- Quotes
- Statistics
- Legal references
- Exam specifications
- "Facts" that sound plausible but are fabricated
Anything you'd stake your professional reputation on, you check. Every time.
Rule 3 — Human always signs off
Under KCSIE and across all school compliance frameworks, a named human is accountable for final content. AI is a drafting tool, never the author of record.
Build this into your team's workflow explicitly: "AI draft → human review → human signoff." No shortcuts.
Rule 4 — Declare use where appropriate
If AI has contributed substantially to a document, consider noting it. Not for every email draft — but for policies, formal communications, and reports, a brief note builds trust and models transparency.
Rule 5 — Keep learning in the loop
The University of Pennsylvania study is the warning. AI that hands over answers undermines learning. AI that hands over time protects it. When you're deploying AI anywhere near pupil learning, ask: Does this preserve the struggle that creates comprehension, or does it short-circuit it?
Getting Started
For teachers — your first 30 minutes
- 1Pick one use case from Part 1. Start with differentiation or report comments — they have the highest time-saving return.
- 2Set up a free ChatGPT or Claude account. Don't use a school-managed tool yet — you want the friction of personal use, because you'll iterate faster.
- 3Try the prompt with real material from your next lesson. Not a test — a real planning task.
- 4Review, refine, save. Keep a note of the prompts that work. You're building your own library.
For heads — your first 30 days
- 1Audit current AI use. Anonymous survey to staff: are you using AI? What for? What's working?
- 2Nominate an AI lead — someone in your SLT or digital team. Their job is to champion, not police.
- 3Align with DfE guidance. Read the June 2025 toolkit and complete the Chartered College training modules.
- 4Draft or update your AI policy using the gap-analysis prompt in Use Case 1. The highest-leverage thing you'll do this term.
- 5Pick a pilot area: one department, one use case, one term. Measure time saved. Report back to SLT.
Frequently Asked Questions
Is this compliant with GDPR?
Yes — provided you follow Rule 1. No personal data in open AI tools, full stop.
Will Ofsted mark us down for using AI?
No. Ofsted's 2026 framework explicitly recognises AI as part of schools' digital strategy. Not using it thoughtfully is now more likely to attract attention than using it.
What if a pupil uses AI to cheat?
This is a real issue, covered in detail in the DfE guidance and JCQ exam guidance. The short answer: assessment design has to evolve. Process-based assessment, in-class drafting, and viva-style checks are the best available responses.
Which tool should we use?
For most use cases, ChatGPT (OpenAI) and Claude (Anthropic) are the leading general-purpose tools. For education-specific tools, Aila (Oak National Academy) is being actively trialled by the EEF in 2026. Don't commit to a platform too early — the market is still settling.
What about pupils with SEND?
AI is genuinely promising for SEND support — adapting material, translating, visualising, scaffolding. But it's not a replacement for specialist provision. Treat it as an amplifier of what your SENCO and support staff are already doing, not a substitute.
How do I know if I'm using it too much?
You're not, yet. The risk in 2026 isn't over-use in schools — it's under-use. If you find yourself outsourcing professional judgement rather than administrative overhead, that's the line to watch for.
About Re-Thrive
Re-Thrive Ventures helps mid-market organisations apply AI to real operational problems. This handbook is free, and will remain free. If your school or trust wants to go further — AI strategy, implementation support, or bespoke training — we'd be happy to talk.
Get in touch: matt@re-thrive.co
Web: re-thrive.co
Version 1.0 · Published April 2026 · This handbook is updated as the DfE guidance evolves.