AI in Classrooms Is No Longer a Future Problem

AI has moved from theory to daily use. According to Tess report on the NEU survey, 76% of respondents now use AI in their daily work, up from 53% last year. Yet 49% said their setting still has no policy for AI use by staff or children.

That gap really matters. When adults use AI well, it can take friction out of the job. Gallups 2025 study found regular users saved an estimated 5.9 hours a week, and many said the quality of their work improved as well. That is a real workload reduction story, and something to be celebrated.

However, the same logic does not carry over to child-facing AI. The question changes as soon as the tool sits inside learning itself. A system that helps an adult draft resources is one thing. A system that answers for a child, nudges their reasoning, or finishes the hard part of a task is something else entirely.  Remember the whole “productive struggle” thing?

The risk reaches beyond just cheating results. Importantly, it touches judgement, patience, effort, and the habit of thinking things through (Remember ”Stop And Think”). Tess coverage of the NEU survey found that 66% of secondary respondents believed critical thinking had declined because of childrens AI use. That should make every responsible adult pause for a moment to “Stop and think” themselves.

If a tool makes work faster but thinking shallower, the trade is too expensive.

There is also a well-being angle. In Education Weeks piece on AI and well-being, the authors argue that the real test is whether AI makes the job more sustainable. That is the right frame for adults. It also highlights why children need tighter boundaries. The same technology that reduces pressure for adults can create dependency for children if it is left open-ended.

Design is key.

The UK government is already acting as if these are two different categories, which they are. In its April 2026 AI tutoring announcement, the Department for Education set out a programme for curriculum-aligned tutoring tools, testing under supervision, with national standards and a focus on disadvantaged children. That is a more serious model than simply dropping a chatbot into a learning environment and hoping for the best, although I would argue that their procurement criteria are misguided, but that’s for another article.

That is the line schools, providers, and platform builders should draw.

Adults can use AI for planning, differentiation, admin, and routine drafting, with review and accountability.

Children should get tools that are tightly bounded, age-appropriate, and designed to support thinking rather than replace it.  And, critically, they must be safe.  We don’t want another Web 2.0-style rabbit hole, thank you very much.

If AI is entering the classroom and homework this fast, we need clarity.

One playbook for adults.

One playbook for children.

Separate rules, separate risks, separate expectations.

Developers and those making the choices about these tools need to understand this critical separation.  Otherwise, teachers would be unnecessarily restricted by tough policy, or children will be exposed to ineffective or even dangerous learning tools.

We’d love to hear your thoughts on this. Find us on the social channels linked at the top of the page.