John Broadbent is a deputy headteacher with 20 years of teaching experience and 10 years in his current post. He has developed and piloted a peer-reviewed AI implementation framework forschools, and delivered training on responsible AI use to NEU members and school leaders.
I’m not a tech evangelist, nowhere near. I’m a Deputy Headteacher with twenty years in the classroom and ten in leadership. I started exploring AI for one reason: the administrative burden was eating into the work that actually matters. Policies to update. Monitoring data to synthesise. Action plans to write. The overhead was stealing time from coaching teachers, having difficult conversations, and being present for pupils – as well as time from my own family.
So I started asking a different question: not should I use AI, but whether it could handle the heavy lifting so I could focus on what requires human judgement.
The answer is yes. But only if you do it right.
The distinction: when to and when not to use AI.
There are two kinds of work in school leadership. There’s overhead – gathering, synthesising, drafting, adapting. It is time-consuming, necessary, but low on professional judgement. And there’s judgement work that requires expertise and professional thinking – deciding what data means in your context, building strategy, coaching staff, personalising communication with families.
AI is genuinely powerful for overhead. It shouldn’t go near judgement.
This matters for professional autonomy. The NEU is right to flag that AI risks devaluing teacher expertise if implemented without care. But the reverse is also true: when AI handles the gathering and organising , it frees you for the work that only you can do. That’s not deskilling – it’s protecting your professional role by clearing the path. AI should expand teacher autonomy, not erode it.
The reality in schools right now
The NEU’s own data tells a stark story. Three quarters of teachers are now using AI for day-to-day work – up from just over half the previous year. Yet nearly half of schools have no AI policy whatsoever. That gap between use and governance is where the risk lies
I saw this first-hand, with teachers already using AI – generic, consumer-facing tools, outside school systems, with no framework around them. Not out of carelessness, but out of good intentions and no guidance.
The risk isn’t AI. It’s AI without oversight.
What actually works – and it's probably already in your school
Most schools don’t realise: you don’t need AI bots or expensive EdTech products. The tools you already have – Microsoft 365 and Google Workspace (for the majority of schools) – contain genuinely capable AI that operates within your secure environment. No procurement process. No new data risk. Just tools you already own, used well.
Copilot (Microsoft 365) and Gemini (Google Workspace) work brilliantly for policy comparison and drafting. Upload your current safeguarding policy, add the latest KCSIE guidance, and ask the AI what’s changed. I watched a delegate complete their entire science policy update in an NEU workshop – work that would normally take considerably longer.
NotebookLM ( part of Google Workspace) goes deeper, cross-referencing multiple sources simultaneously. I use it to synthesise monitoring data from lesson observations, pupil voice, assessment results, and research – all at once. Ask the right questions and it identifies patterns you’d miss manually.
The prompt makes the difference
Vague prompts produce vague outputs. Specific, structured prompts produce something genuinely useful. A few examples:
“Here’s our 2024 safeguarding policy. Here’s the new KCSIE guidance. What’s changed? What new areas should we emphasise?”
“Here are this term’s lesson observation notes, pupil voice feedback, and assessment data. What are our three most important literacy priorities for next year?”
“Draft a letter to parents about [topic]. Friendly but professional tone. These key points only. No pupil names.”
In each case, you’re giving the AI the raw material and a clear task. You’re still making the decision. The AI is doing the organising.
How to create oversight
You don’t need bureaucracy. You need clarity. Building and piloting a peer-reviewed AI implementation framework in my own school taught me that these foundations are what actually matter:
Find out what staff are using – without judgment. Most aren’t doing anything wrong; they just haven’t thought through the implications. Point them to the tools they already have. Set clear boundaries: no pupil names, no identifying data, no safeguarding information in any AI input. Build the skill – show staff how to prompt well, check outputs critically, and understand why human verification is non-negotiable. And consult before you implement. The NEU’s position is clear on this: staff involved from the start become allies.
Always keep humans in charge. AI processes information. You make the decision.
Why this matters
The workload problem in schools is real. AI can help – with the overhead, the synthesis, the drafting. But we must be careful that AI does not intensify work, displace professional judgement, or paper over funding pressures. A framework that’s clear about what AI is for – and what it isn’t – is the only way to make sure it serves teachers rather than diminishes them.
I gained time back. Time with my staff, my pupils, my family. But the reason it worked isn’t the technology. It’s the approach: pedagogy first, guardrails as support, staff who understand why not just what they’re told.
The question is not: should we use AI? It is: how do we use it well?