AI in the classroom: an honest guide for primary teachers

There is a lot of noise about AI in education. Some of it is enthusiastic to the point of recklessness; some of it is anxious to the point of paralysis. Most primary teachers are somewhere in between — curious, a little uncertain, and mostly too busy to read 3,000-word think pieces about the future of learning.

Here is a practical, honest guide to what AI tools can and can't do for primary school teachers right now.

What AI is genuinely useful for

Drafting. AI is very good at producing a workable first draft of something that follows a pattern — a report comment, a parent email, a lesson plan outline, a set of differentiated tasks. The draft is never perfect, but starting from a draft is faster than starting from nothing, and the quality of AI drafts has improved significantly.

Generating options. When you're stuck on how to phrase something, AI can give you five different approaches in seconds. You pick the one that fits, adjust it to sound like you, and move on. This is different from using AI to write something wholesale — it's using it as a resource when you're stuck.

Saving repetitive work. Tasks that follow a clear pattern — generating exit tickets, writing reading record comments, producing similar letters for multiple families — are well-suited to AI assistance. The skill and judgement are still yours; the repetitive generation work is offloaded.

Planning at scale. Generating a unit plan outline, a sequence of lesson objectives, or a range of activities for a topic is something AI does usefully well. You still make the professional judgements about what is right for your class; AI can give you the scaffold to work from.

What AI is not useful for

Professional judgements. Whether a child is ready to move on, how to handle a sensitive parent conversation, what a piece of work reveals about a child's understanding — these require human knowledge, context, and relationship. AI cannot provide them.

Anything requiring real knowledge of a specific child. AI can draft a report comment for “a Year 4 child who is strong in maths but lacks confidence in writing.” It cannot know that Oliver has been trying harder since his friend moved schools, or that Maya's reading has suddenly taken off since she discovered the book series she loves. Those things have to come from you.

Making things simpler in ways that matter. AI can make the mechanical work faster. It cannot make the job easier in the ways that actually count.

The GDPR question

Any AI tool that processes pupil data — names, year groups, academic information — needs to be GDPR compliant. Before using any AI tool with information that relates to a specific child, check: where is the data stored? Who has access to it? Does it get used to train AI models?

These are not paranoid questions — they are the right professional questions to ask.