The Minutes AI Gives Back: Are Schools Spending Them on Better Curriculum Thinking?
Photo: OpenAI image generation tramite ChatGPT; prompt, composizione editoriale e testi di Alfredo Vassalluzzo, Public domain, via Wikimedia Commons
Generative AI has moved from novelty to routine in many UK schools. In 2026, teachers use it to draft lesson activities, generate examples, adapt texts, produce quizzes and format resources. The workload case is easy to see: a task that once took an hour can sometimes be reduced to minutes. But the more important question is not whether AI saves time. It is what happens to that time, and to the professional thinking it once demanded.
Saved time is not automatically reinvested. It can be absorbed by cover, email or the next administrative task. It can also change the shape of planning: a teacher may move from composing a sequence of learning to selecting from generated options. That shift is not necessarily bad, but it is pedagogically significant. If AI produces a plausible lesson and the teacher has not had to resolve the underlying curriculum problem, the plan may look efficient while the thinking stays shallow.
Where AI genuinely helps
AI tends to be most useful when the teacher already knows the destination. It can generate multiple examples of a concept, vary the readability of a text, suggest common misconceptions, or produce low-stakes retrieval questions. It can help a department build a bank of practice questions and then subject them to human review. For workload, its strongest contribution is often in the production of variants rather than in the invention of the curriculum.
It also helps when the task is bounded and the criteria are explicit. A prompt that asks for five examples of a particular grammatical structure is different from a prompt that asks for a lesson on grammar. The first leaves the teacher in control of sequence and purpose. The second invites the model to make curriculum decisions that belong to the teacher.
Where it can flatten pedagogy
AI can narrow thinking when it is used to bypass the productive struggle of planning. If a teacher never has to decide why a particular example comes first, how an analogy might mislead, or what evidence will show that pupils have understood, the plan may be tidy but the teacher's subject knowledge is not being exercised. Over time, that matters.
There is also a risk of homogenisation. Generated resources often default to conventional formats: a starter, some explanation, a worksheet, a plenary. That structure can be useful, but it can crowd out other approaches — extended writing, structured discussion, practical work, retrieval practice designed around a specific misconception. The model does not know the class. It does not know that Year 8 have just returned from a difficult lunchtime, or that a particular pupil needs the text chunked differently.
A further concern is curriculum coherence. Individual AI-generated lessons can be strong while the sequence is weak. If planning is fragmented across a department, pupils may meet the same concept in three different ways without anyone noticing. AI does not automatically create the vertical thinking that curriculum design requires.
A simple audit for departments
Before celebrating time saved, departments should ask where it went. A short audit can be done in a department meeting:
| Planning task | AI used? | Time saved | Time reinvested in... | Teacher judgement visible? |
|---|---|---|---|---|
| Example generation | Yes/No | Minutes | Adaptation, assessment, subject reading | Yes/No |
| Text adaptation | Yes/No | Minutes | Checking accessibility and challenge | Yes/No |
| Quiz production | Yes/No | Minutes | Reviewing distractors and misconceptions | Yes/No |
| Lesson sequencing | Yes/No | Minutes | Curriculum coherence, prior knowledge | Yes/No |
The audit is not a compliance tool. Its purpose is to make the trade-off visible. If AI saves time on example generation but that time goes straight into cover, the workload benefit is real but fragile. If it saves time and the department uses it to review assessment items together, the benefit is more likely to improve teaching.
Three department protocols
-
Keep the curriculum question human. Agree that AI may draft resources, but the sequence of learning, the choice of core knowledge and the assessment criteria are decided by teachers. A simple rule helps: no generated resource is used until someone can explain why it is in that position in the sequence.
-
Review for subject accuracy and pedagogical fit. Build a five-minute check into resource creation: Is the subject content correct? Does the example reveal or obscure the key idea? Does the task require the thinking we want? Does it fit this class? This is not a bureaucratic hurdle; it is the point at which teacher judgement is exercised.
-
Protect the time saved for professional purpose. Departments should name one or two uses for AI-generated time — for example, collaborative review of assessment questions, subject reading, or planning a challenging sequence. Without that decision, saved time will be reabsorbed. Senior leaders can support this by protecting department meeting time and resisting the urge to fill it with announcements.
Keeping judgement central
The danger is not that AI writes lessons. The danger is that it quietly changes what teachers attend to. If planning becomes primarily a process of selecting and editing generated content, the curriculum may become more uniform and less thoughtfully sequenced. If it becomes a process of using AI for production while teachers retain responsibility for design, the workload gain can be genuine.
In 2026, the most useful question for school leaders is not 'Are we using AI?' but 'What are teachers doing with the time it returns?' The answer should be visible in department meetings, in the quality of assessment, and in the curriculum conversation. AI can save minutes. It cannot decide what those minutes are for.
