Dear SLT: what your teachers will never say in a meeting
workloadteacher retentionschool leadershipAI policySLT

Dear SLT: what your teachers will never say in a meeting

By Geoff Waugh8 min read

Nobody resigns in a meeting. They resign in a letter that says "family reasons", long after the decision was actually made.

If you lead a school, the most important thing your staff think about their workload is probably the thing you never hear. They will not raise it in briefing. They will not tell their line manager. They will not put their name to it in the staff survey, because they do not want to sound ungrateful for a job they need. They just quietly lay plans to leave.

They mark on weekends. They mark in the PPA and the free periods they needed in order to decompress from a difficult class. On Monday you say "how are you, hope you had a lovely weekend", and they smile and say it was great, thank you. Then one day they hand you a letter that mentions family reasons or a new challenge, and tell you that working here was a privilege, with a reference request in sight.

I am a teacher. I have done the Sunday night marking and smiled through Monday exactly as described. This is not a complaint about leaders. Running a school is a job I genuinely cannot imagine doing, and the people I have worked for could have applied themselves to banking and made a great deal more money. They chose education deliberately. This is a letter about a blind spot, written because the people best placed to tell you have every reason to keep quiet.

The numbers your staff will not say out loud

The data says what the teachers do not.

In the DfE's own Working Lives of Teachers and Leaders survey, published in November 2025, 29% of teachers and leaders said they were considering leaving the English state school sector within 12 months, for reasons other than retirement. Of those, 89% cited high workload. Run that fraction down your staff list if it feels abstract. It is not a national statistic. It is happening in your school, now, among the people whose commitment you admire.

The figure that should stop you is this one. Of teachers who have already left, 78% say their workload is acceptable. Of those still teaching, 26% do. The leavers who are still in work average a 34.8 hour week. The ones who stayed average 48.2.

The work is not unbearable everywhere. It is unbearable here. That is not "how it has to be". That is a choice the sector keeps making.

And this is not staff against SLT. In Education Support's Teacher Wellbeing Index 2025, 86% of senior leaders say they are stressed at work, and 26% have considered leaving their job over the pressure on their own mental health and gone as far as actively looking. Ask the ones who are looking why, and 76% point at the sheer volume of work. Not pupils, not parents, not accountability. The volume. You are drowning in the same water your staff are.

People are going, too. Around 39,000 full time equivalent teachers left state schools in 2024/25, and roughly 40,800 the year before, on the DfE's own school workforce figures. Over three years that is well past 110,000 people. Every one of them is a recruitment round, an induction, and a class that loses continuity. This is a retention crisis, not an HR complaint.

Your staff have already found a workaround

Teachers have not waited for permission.

A YouGov survey for Bett UK, carried out in June 2026 across 1,033 UK teachers, found that 80% now use AI in their work, with weekly use doubling from 26% to 52% in a single year. Just over half said it had reduced their workload. Only around a third said they were actually working fewer hours as a result. Individual teachers improvising with chatbots is not a workload strategy. It is a thousand private experiments with your school's data.

Much of it happens out of sight. When BCS asked in December 2024, only 36% of teachers using AI for lesson preparation had told their senior leadership team. Meanwhile Browne Jacobson's survey of 200 school leaders found just 8% felt prepared to implement AI, and three quarters said their organisation did not have the expertise.

That gap is dangerous right now. DfE guidance is clear: "it is recommended that personal data is not used in generative AI tools", because anything typed in may be stored and learned from. But a tired teacher with 30 essays and a Wednesday deadline is not rereading DfE guidance at 10pm. If you have not given staff a safe tool for a job, assume some of them are using an unsafe one. That is not their failure. Choosing tools was never their job.

Five things you can do before Christmas

None of this needs a three year strategy or a new budget line.

Publish an approved tool list. Name what staff can use, and for what. TeachMate or Oak National Academy's Aila for lesson resources, DeepMark for marking, Copilot inside the licence you are probably already paying for. One page, reviewed termly.

Run one hour of practical training. Not a lecture on the dangers. Show staff how to draft a worksheet, adapt a reading for different needs, or summarise a long email chain. In Twinkl's survey of over 6,500 teachers, 76% said they had not had adequate training or guidance.

Write a one line data rule. For example: no pupil names, candidate numbers or identifiable work in any tool that is not on the approved list. Clear beats comprehensive. Expecting that no work will touch AI is wishful thinking. The cat is out of the bag.

Take the friction out of data drops. Grades still have to land in SIMS or Arbor, and somebody still has to judge the borderline cases. But a well built template that converts marks to grades and lines the columns up for copy and paste turns hours into minutes. Ask your data manager which three steps they repeat most.

Ask the question anonymously. One question, every half term: how many hours did you work at home last week? You will learn more from that than from any staff voice meeting.

Why we built DeepMark

Full disclosure. I co-founded a company in this space, and yes, I would like you to look at it. I think it has earned your time, and I think your teaching staff have earned it too.

DeepMark is a digital marking environment for departments. Teachers mark scanned scripts, everything is stored in one place, marking can be split by question the way exam boards do it, and heads of department can see across every class without chasing spreadsheets. Departments choose how much AI is involved, including none at all, and a teacher signs off every single question.

That is not why we built it, though. We built it because marking is the part of teaching that eats weekends, and nobody chooses to spend a Sunday holding a mark scheme in their head for the thirtieth script. If giving heads of department the oversight they need is how teachers get their weekends back, that is a trade we are happy to make.

Worth noting from that same Bett survey: only 8% of teachers currently use AI for marking. The single biggest sink of teacher time is the one bit almost nobody has been given a safe way to do.

You do not have to use us. Please use something. Nobody complaining is not the same as nothing being broken.

Geoff Waugh, teacher and co founder, DeepMark

Try DeepMark →

Sources

See it live

This is DeepMark marking a real script

Scroll through as it marks. Feedback and annotations appear as they land. Click anything, highlight, edit a mark to get a feel for how it works.

Loading the marking editor…