Teacher workload is the problem the profession has been trying to solve for a decade, and AI is the first tool that genuinely dents it. It can plan, differentiate and draft feedback in minutes. It can also leak pupil data and mark work it should not, which is why training and policy come first.
The summary
Yes, teachers can claw back hours a week with AI on lesson planning, differentiation, resources and first-draft feedback. Tools like Google Gemini for Education, Microsoft Copilot and purpose-built assistants such as TeachMateAI fit school life. The risks are specific: pupil personal data, safeguarding, accountability for marking, and over-reliance. The DfE and KCSIE set the frame. Train inside it and the workload win is real and safe.
What can teachers actually use AI for?
The best AI use in teaching is preparation, not judgement. It gets you to a strong first draft of a lesson or a resource, leaving the teaching and the marking decisions where they belong, with the teacher.
| Everyday task | How AI helps |
|---|---|
| Lesson and unit planning | Drafts plans, objectives and activities you shape to your class. |
| Differentiation | Produces the same task at several levels, or scaffolds for pupils who need it. |
| Resources and worksheets | Generates first-draft worksheets, quizzes and knowledge organisers. |
| First-draft feedback | Turns your comments into pupil-friendly feedback you check and own. |
| Admin and communication | Drafts parent emails, reports and newsletters in a consistent tone. |
Which AI tools suit teachers?
In a school, the data question decides the tool. Anything involving pupils belongs in an education-grade, school-controlled environment.
| Tool | What it is used for | What to watch |
|---|---|---|
| Google Gemini for Education | Planning, resources and admin inside a school Google Workspace | Use the education tenant, not a personal account, for anything school-related. |
| Microsoft Copilot | Planning and admin inside a school Microsoft tenant | Governance settings decide what it can see; set them first. |
| TeachMateAI or similar | Teacher-specific tasks like plans, reports and letters | Still a first draft; the professional judgement is yours. |
| Diffit | Levelled reading resources from a text or topic | Check reading age and accuracy for your class. |
| ChatGPT or Claude | General planning and drafting with no pupil data | Never enter pupil personal or sensitive information. |
Three frameworks for training teachers on AI
A one-size course does not fit a staffroom, because an English teacher and a SENCo have different needs. These three frameworks make training land.
1. Role-based upskilling
Role-based upskilling trains departments on what they actually do. A core subject team might focus on differentiation and feedback; a pastoral team on communication and reporting. The tasks come from the timetable, not a generic list.
Act as an expert trainer for schools. Analyse the weekly workload of our Key Stage 3 English department and identify the top 5 tasks where generative AI can cut preparation time by 30%. Give a 4-week practical training outline with tool recommendations, hands-on exercises using our real schemes of work, and simple measures to track adoption and workload impact.2. Managing change and resistance
Teachers are, rightly, protective of professional judgement and wary of anything that looks like deskilling. The change framework meets that by framing AI as a way to give time back to teaching, and by teaching delegation rather than clever prompting.
We are introducing AI tools across our school but staff worry it will deskill teaching or add to workload. Design a change-management plan and a 1-hour staff-meeting workshop. Focus on teaching staff to delegate preparation tasks to AI to protect teaching time, rather than technical prompting, and outline how to handle common objections from teachers.3. Security, governance and ethics
Governance in a school is safeguarding first. Staff need to know that pupil data never goes into a public tool, that AI feedback and marking remain the teacher's responsibility, and how to check AI output for accuracy and bias before it reaches a pupil or parent.
Create a clear, non-technical AI acceptable-use policy and a short quiz for a school staff training module, aligned to DfE guidance and KCSIE. Teach the difference between public and education-grade AI, what pupil data must never be entered, how a teacher stays accountable for marking, and how to fact-check and check for bias before AI output reaches pupils or parents.Example AI prompts for teachers
Adapt these to your subject and class. None of them replaces your professional judgement on the day.
| Task | Prompt you can adapt |
|---|---|
| Lesson plan | Draft a one-hour lesson plan on this topic for a mixed-ability Year 8 class, with objectives, a starter, a main activity and a plenary, in UK English. |
| Differentiation | Take the worksheet below and produce three versions: a supported version, a core version and a stretch version, keeping the same learning objective. |
| Feedback | Turn my marking notes below into clear, encouraging feedback for a pupil, focused on two things they did well and two next steps. |
| Knowledge organiser | Create a one-page knowledge organiser for this topic with key vocabulary, definitions and five recall questions. |
| Parent email | Draft a warm, professional email to a parent about their child's progress, using the points below, under 200 words. |
What are the risks of AI for teachers?
These are the exact points a good school training session is built around.
| Risk | Why it matters | How training handles it |
|---|---|---|
| Pupil data and safeguarding | Entering pupil personal or sensitive data into a public tool breaches data law and safeguarding duties. | Education-grade tools only, with pupil data kept out of anything public. |
| Accountability for marking | A teacher remains responsible for any assessment, whoever drafted it. | AI drafts feedback; the teacher checks, edits and owns it. |
| Accuracy and bias | AI content can be wrong or skewed, and pupils will not know. | Check every resource for accuracy and bias before use. |
| Over-reliance | Leaning on AI can thin out subject expertise over time. | Keep the teacher's judgement central; AI supports, it does not decide. |
How should a school roll out AI?
Schools that get value start with one department and the tasks that eat the most time, prove the workload saving, and widen with a policy already agreed.
| Step | What it involves |
|---|---|
| 1. Set the policy first | Agree an AI use policy aligned to DfE guidance and KCSIE before classroom use. |
| 2. Start with one department | Pick a team and the tasks that cost the most preparation time. |
| 3. Train on real schemes of work | Run hands-on sessions using your own curriculum, with accuracy checks built in. |
| 4. Measure workload and widen | Track time saved and roll it out to the next department. |
Want to give your teachers their evenings back?
We train school teams on the AI that cuts planning and marking time, inside the DfE and safeguarding frame, with pupil data kept safe. Practical, subject-aware and workload-focused.
Take Our Free 2 Minute AI Skills Assessment
Harry Lang runs The Oxford AI School, teaching teams and businesses the basics of AI in a fun, easy and practical way. He can be reached at Harry@TheOxfordAISchool.com.
AI Training for Teachers: FAQ
Can teachers use ChatGPT for planning and marking?
For planning and first-draft resources with no pupil data, yes. For anything involving pupils, use an education-grade, school-controlled tool and keep pupil personal data out of public tools. Marking can be drafted by AI, but the teacher checks, edits and remains accountable for it.
What does the DfE say about AI in schools?
The DfE supports using generative AI to reduce workload while making clear that schools must protect pupil data, follow safeguarding duties under KCSIE, and keep teachers accountable for outcomes. The technology is encouraged inside those limits, not banned.
Will AI reduce teacher workload or add to it?
Used on the right tasks, it reduces it, mainly on planning, differentiation, resources and first-draft feedback. Poorly introduced, without training or policy, it can add friction. That is why role-based training and a clear policy matter.
What is the biggest AI risk for a school?
Pupil data and safeguarding. Entering pupil information into a public tool is the main danger, followed by accepting inaccurate or biased content. Both are handled by policy and training before any classroom use.
How long does AI training for teachers take?
A focused twilight or INSET session on real schemes of work gets a department going, and most teachers save time within a couple of weeks of practising. Training built around your own curriculum beats a generic course every time.