The Hidden Cost of AI in Schools

Much has been written about what artificial intelligence might do to pupils. Will it undermine homework? Will it fuel plagiarism? Will it widen gaps?

These are important questions. But they may not be the most important ones.

AI may not erode pupil learning first. It may erode teacher expertise.

I am not anti-AI. I lead digital strategy in my school. I use AI. I believe it can reduce workload and improve efficiency. We absolutely need that. Teachers are exhausted, and any serious conversation about AI must acknowledge that reality.

Reducing workload is not the same as outsourcing professional thinking.

Planning is not just preparation. It is professional development.

When teachers plan lessons from scratch, they wrestle with the specification. They interrogate the sequencing. They decide what to leave out and what to emphasise. They anticipate misconceptions. In doing so, they deepen their own subject knowledge.

Increasingly, new teachers enter a culture of ready-made lessons. Shared drives. Downloaded schemes. Fully formed PowerPoints. Now AI adds another layer: generate me a lesson on X, with starter, main task and plenary.

It looks efficient. It feels efficient.

But something subtle is lost. The slow, deliberate act of thinking through a topic. The productive struggle that sharpens understanding. The habit of asking: Why is this concept hard? Where will they go wrong?

Yes, it is time-consuming. Yes, we must streamline it. But marking is also where teachers learn. It is where we see the almost-right answer. The half-understood definition. The tiny misconception that reveals a deeper misunderstanding of the curriculum.

Exam board marking has increased my own understanding of what is really being rewarded. You see patterns. You notice how small wording choices change marks. You internalise the standard. That knowledge feeds directly back into teaching.

If AI marks everything, teachers risk losing that feedback loop. We may reduce workload but also reduce professional calibration.

There is also a pacing issue.

PowerPoint quietly accelerated lessons years ago. Everything pre-built. Every explanation polished. The temptation was to move too quickly because the slides were ready.

AI risks amplifying that effect. If explanations, models and worked examples appear instantly, the messy, step-by-step modelling can disappear. Lessons become smoother but shallower.

None of this means we should reject AI. It means we must use it with intent.

Automatic marking of low-stakes homework can be helpful. AI-generated quizzes can support retrieval. Drafting administrative emails or formatting resources is sensible.

But assessments and exams? Those are professional judgements. There are moments where teachers refine their understanding of pupils and curriculum. Outsourcing them wholesale would be short-sighted.

The question is not whether AI reduces workload. It clearly can.

The question is whether it replaces professional thinking or protects it.

If we allow AI to become the default planner, the default explainer and the default marker, we risk creating teachers who are increasingly managers of systems rather than experts in subjects.

Teaching is not content delivery. It is interpretation, diagnosis and adaptation.

Efficiency matters. But expertise matters more.

Use AI to remove friction. Do not use it to remove thinking.

Because once professional knowledge atrophies, rebuilding it will be far harder than generating the next lesson plan.