The debate about AI in education is often framed around cheating by pupils or teachers’ jobs being replaced.

Both are compelling debate points and represent clear risks. Yet I think there is something more nuanced and more worrying about AI that we are overlooking: how it could erode teacher and pupil judgement if it is introduced too early and without the time for mastery to develop.

For example, AI is exceptionally good at expanding capacity. It can generate lesson materials within seconds, draft answers, summarise texts and streamline marking.

AI outpacing professional judgement

For busy teachers, that efficiency is attractive. However, the problem is not increased output but what happens when use of AI outpaces the development of professional judgement.

Because professional judgement is not acquired instantly. It develops through trial, error and reflection: when a lesson does not quite land, when an activity turns out to be too obvious or when students reach answers too quickly without having to justify their thinking.

These moments are not inefficiencies; they are how teachers learn to diagnose problems and refine their craft. It can take years to hone this.

AI can accelerate iteration, but it cannot replace those diagnostic instincts. An experienced teacher might use it to generate several versions of a case study, then deliberately remove the easy option, add competing priorities or introduce ambiguity to raise the cognitive demand.

Refinement is the key

The value of AI lies not in generation but in refinement. AI does not replace pedagogical judgement; it rewards people who already have it.

If AI is used before that judgement is secure, however, it risks creating a different pattern.

Lessons may appear slick, resources may look well-structured and workload may feel lighter. But the teacher may be doing less of the slow thinking that builds long-term expertise: analysing why a task fell flat, spotting subtle misconceptions or deciding when to let students struggle productively. Over time, that matters.

If judgement weakens, the consequences are not dramatic at first. Standards do not collapse overnight. Instead, quality drifts. Tasks become safer, explanations become flatter and professional confidence increasingly rests on what the tool produces rather than on internal conviction.

The loss is subtle, but over years it reshapes how expertise is built and sustained.

Earning the right to use AI

This is why AI should be understood as something that must be earned before it can be used to augment a teacher’s work.

This earned augmentation does not mean bans or blanket restrictions. But it does mean being judicious about when and how AI can be used and sequencing it with thought into professional development.

For example, early career teachers need protected time to plan, mark and reflect in full, building the instincts that allow them to recognise quality and respond to it.

That may mean resisting the temptation to use AI as a shortcut during the very phase when judgement is still forming. Introducing heavy reliance on AI at that stage may ease workload in the short term, but it could also limit the formation of those instincts.

More experienced teachers, by contrast, may be better placed to use AI extensively because they can evaluate and reshape what it produces, using it as a multiplier of expertise rather than a substitute for it.

The same rules for students

Schools do not need a detailed policy overnight, nor a rigid ban on AI use. But even recognising that timing matters, and that professional judgement should be formed before capacity is heavily amplified, would be a meaningful shift.

The same principle applies to students. When learners use AI to structure arguments or generate ideas before they can do so independently, they may produce work that appears fluent but is built on fragile foundations

If the aim of education is to develop independent thinkers rather than competent imitators, then access to augmentation should follow the development of understanding, not replace it.

This is not an anti-AI argument. Used at the right time, AI can reduce friction and protect energy for higher-value decisions.

But if we care about professional judgement in teachers and intellectual independence in students, we cannot treat augmentation as neutral or inevitable.

The question is not whether AI belongs in education. It is whether we are willing to decide when it should be introduced and what must be built first.

Richard Watts is a teacher of business at Richard Taunton Sixth Form College in Southampton

You can now get the UK’s most-trusted source of education news in a mobile app. Get Tes magazine on iOS and on Android