AI won’t replace teachers, but it’s already taking over parts of the work around teaching, like first drafts of lesson plans, quizzes and reports. What it can’t take is the judgement and the relationship in the room. So the useful question for leaders is what to do with the time it gives back.
Say it’s Sunday evening and you’ve got a week of Year 5 lessons to plan. An AI tool will happily draft Monday’s maths. It can’t tell you who went quiet on Friday, or that this class switches off straight after lunch. That part’s still yours, and it’s the part I’d protect first.
What AI is already doing in teachers’ weeks
Most teachers already use it, and they use it on the preparation around lessons far more than on the teaching itself. Teacher Tapp’s questions on AI in late 2025 drew answers from roughly 8,000 to 10,000 teachers in England, weighted to match the national workforce. Only 12% had never used it for schoolwork (Chartered College of Teaching, 2026).
Look at what they use it for. Among teachers who’d tried it, 54% used it for lesson planning, 50% for quizzes and 46% for marking. Just 11% used it to deliver lessons. So AI is doing the work around the lesson, and the teacher is still the one in front of the class.
How much time AI really saves
Less than the hype suggests, but it’s real. In a trial published in October 2026, AI saved primary teachers in England about 49 minutes a week on lesson preparation, and the quality was no worse.
The National Foundation for Educational Research ran it for the Education Endowment Foundation. It randomly split 464 teachers of Years 3 to 6, in 108 schools, into two groups for ten weeks of the 2025 autumn term. One group used Aila, Oak National Academy’s AI lesson assistant. Their preparation time came out at 76% of the other group’s, about 2.5 hours a week against 3.3. Independent teacher reviewers rated their resources no worse (Education Endowment Foundation, 2026). Three in four of the teachers had already used AI to plan lessons before the trial began.
An earlier EEF trial with secondary science teachers pointed the same way, with planning time falling to 69% of the comparison group’s (Education Endowment Foundation, 2024).
In the 2026 trial, teachers used Aila for only 7% of their lessons. Most of them still adapted what it produced, and many felt it needed changes before it was ready for the classroom.
Now set 49 minutes against the whole week. Teachers in England, counting part-timers, reported working 46.9 hours a week on average in 2025 (Department for Education, 2025). The trial measured one task, and that’s where the saving sits.
When I wrote about AI as a teaching assistant in 2025, the time savings were mostly an estimate. Now there’s a randomised trial, and it gives a smaller, firmer answer: real minutes, on the tasks around teaching.
Why AI can’t replace teachers
The parts of teaching that matter most to pupils, the judgement and the relationship, can’t be written down as a list of tasks. A list is exactly what AI is good at.
That’s my list test, and I’d apply it to any job. If you can write everything you do down as a list, AI is coming for that list. Run it on a teacher’s week and the tasks and the job come apart. Plenty of tasks go on the list, like the quiz, the worksheet and the first draft of a report. The job itself doesn’t fit on it. You can’t write down how to win back a Year 9 who’s decided maths isn’t for him.
The Department for Education says much the same in its guidance for schools in England. Generative AI “cannot replace the judgement and deep subject knowledge of a human expert” (Department for Education, 2025). It also “should not replace the valuable relationship between teachers and pupils”.
There’s evidence behind that relationship, too. A 2025 review in Psychological Bulletin pulled together 26 meta-analyses covering about 2.64 million children, from pre-school to the end of secondary. Positive teacher and pupil relationships were linked with achievement, motivation, behaviour, belonging and wellbeing (Emslander and colleagues, 2025). These are links rather than proof of cause, but the pattern is large and it’s consistent.
The five human intelligences AI can’t automate
Curiosity, Creativity, Critical Thinking, Adaptive Leadership and Emotional Intelligence are the five I’d protect in every classroom. AI can’t automate any of them, and growing them is the real job.
This is my framework, not a lab finding. I’ve hired and trained more than 1,000 people, and when I hire, these five are what I’m really testing for.
Curiosity looks like the question a pupil asks that isn’t on the worksheet. It’s the B in BRAVE, the framework from my TEDxSwansea talk: Build Curiosity. You see Creativity when a teacher rewrites the plan at 10.15 because it isn’t landing.
Critical Thinking means spotting when the AI is confidently wrong. The DfE’s guidance, updated in 2025, warns that these tools can produce “inaccurate or false information presented as fact”. I’ve written about teaching pupils to question AI rather than trust it.
Adaptive Leadership is the teacher who changes course when the room changes. And Emotional Intelligence is noticing who went quiet on Friday, and knowing what to do about it on Monday.
I come back to one line on this.
“AI masters what already exists. But humans are the architects of what’s next.”
Pick one of the five and find where it shows up in next week’s timetable. If you can’t find it, that’s your first job.
Are teachers at risk from AI?
Some of their tasks are, but the bigger risk to teaching right now is teachers leaving, and workload and stress are the reasons they give most.
The DfE’s 2025 survey of 10,808 teachers and leaders in England found 29% were considering leaving the state sector within a year, for reasons other than retirement. Of those, 89% named high workload, and 89% named stress or poor wellbeing (Department for Education, 2025).
I’ve said it before and I’ll keep saying it. “Teachers haven’t lost their spark to teach. They’re drowning in paperwork and admin.”
So I’d point AI at the paperwork, never at the teacher, and US policy is heading the same way. The Department of Education’s proposed grant priority on AI, published in 2025, includes using it to “reduce administrative burdens” (US Department of Education, 2025). If your staff room is still stuck on whether to allow AI, I’ve written about how fear of AI in classrooms holds schools back.
What to do with the time AI gives back
Decide where it goes, in writing, before it quietly fills up with more admin. I’d do it in four steps this half term.
- List a teacher’s week. Ask two or three teachers to write down everything they did last week, task by task. Then run the list test on each line and mark what AI could draft.
- Name what AI may draft, and who signs it off. Quizzes, first-draft plans and report comments are fair game. The DfE says anything AI produces needs “critical judgement to check for appropriateness and accuracy” (Department for Education, 2025). It stays the responsibility of the professional and the school.
- Write down where the minutes go. Feedback conversations, small-group work, ten minutes with the pupil who needs it, or leaving at a sensible hour. If you don’t name it, more admin will take it.
- Put it in your guidance, and review it at half term. Gallup’s early 2026 survey of 2,069 US public school teachers found only 18% had formal guidance on AI from their administrators (Gallup, 2026). Don’t leave your staff guessing.
Say you lead a primary school with 20 teachers. If each one saved what the trial teachers saved, that’s more than 16 hours of teacher time a week across your staff. Start step one on Monday, with one teacher and one week of tasks.
How Colin helps
I’ve spent about 17 years in AI, XR and emerging tech, and close to two decades in human behaviour and neuroscience. So I sit on both sides of this question.
We built Illuminate XR, an immersive AI and VR learning platform for schools, and I’m its co-founder and co-CEO. That means I’ve a stake in schools using this technology well, and you should know it. BRAVE, the education framework from my TEDx talk, is available to educators at no cost.
If you’d like to talk about a keynote or advisory work for your school, university or business, get in touch.
Related questions
Will AI replace teachers in the UK?
Not on current policy. The Department for Education’s guidance for schools in England, updated in 2025, says generative AI can’t replace a human expert’s judgement. It also says AI shouldn’t replace the relationship between teachers and pupils. The guidance treats AI as a way to cut workload, with teachers checking what it produces.
Can AI replace teachers in the future?
Nobody can promise what AI will do in ten years, and I’d be wary of anyone who does. My list test still sorts it. Whatever can be written down as a list of tasks will keep moving to AI. The relationship and the judgement in the room don’t fit on that list.
Should AI replace teachers?
No. Pupils need an adult who knows them. A 2025 review of 26 meta-analyses in Psychological Bulletin linked strong teacher and pupil relationships with achievement and wellbeing. Use AI to give that adult more time with the class.
Is it worth becoming a teacher in 2026?
I think so, if it’s the human side of the job that draws you, because that’s the part AI leaves alone. Go in clear-eyed about workload. Only 26% of teachers and leaders in England called theirs acceptable in 2025, though that’s up from 22% a year earlier (Department for Education, 2025).