Ask a staff room whether anyone has used an AI tool to draft a lesson plan and the answer is increasingly yes. Ask whether the school has a written position on it and the answer is usually no.
That gap is where the problems live. The technology question has largely settled itself in practice; the governance question has not been asked, and in most schools nobody has been made responsible for asking it.
Where the tools are genuinely useful
Planning scaffolds, differentiated versions of the same worksheet, and first-draft comments on routine work are the uses teachers report most often. All three share a shape: the tool produces a draft, and a professional edits it.
That is a reasonable division of labour, and it is worth defending. A teacher who gets an hour back from formatting is a teacher with an hour for the class in front of them.
Where it goes wrong
It goes wrong when the draft is not edited. When student work is pasted into a tool without any thought about whose data it is. When a generated comment reaches a parent without a teacher having read it.
None of those are technology failures. They are the predictable result of having no stated rule, in an environment where every individual is making the call alone and under time pressure.
Settle the data question first: student names, marks and personal circumstances do not go into a general-purpose tool. Everything else can be worked out as the tools change, and they will keep changing.
The data rule is the one that matters
Of everything a school could write down, one rule does most of the work: identifiable student information stays out of general-purpose tools.
Not names, not marks, not disciplinary or medical detail, not anything a parent would be surprised to learn had been typed into a commercial service. This is not a technical restriction — it is the same duty of care that already governs what goes in a staff WhatsApp group.
What a usable policy contains
A short document is more effective than a long one, because a long one is not read. Four questions, answered plainly, cover most of the ground:
| Question | A workable answer |
|---|---|
| What may go into a tool? | Topics, objectives, generic examples, your own words |
| What may never? | Student names, marks, personal or medical circumstances |
| Who checks output before it reaches a student or parent? | The teacher whose name is on it. Always |
| What do I do when unsure? | Ask a named person, before rather than after |
The marking question
Using a tool to draft feedback on routine work is where most teachers start, and it is defensible as long as the teacher reads what goes out. The trouble is that generated feedback is fluent, which makes it easy to skim and approve.
Fluent and wrong is a worse combination than clumsy and right. A comment that praises reasoning the student did not show, or corrects an error they did not make, tells a parent that nobody read the work — and they will notice, because a student who was there knows what they wrote.
The practical safeguard is small: read the piece before you read the generated comment, not afterwards. Once you have seen the suggestion it is very hard to assess the work independently.
Who owns the decision
In most schools the honest answer is nobody, which is how a policy vacuum persists in an institution that has policies for everything else. Somebody has to be the person a teacher asks at half past four when they are unsure, and that person has to have an answer.
It does not need to be a senior leader, and in schools where it works it often is not. It needs to be someone who has actually used the tools, is trusted by colleagues, and has authority to say no.
Assessment is a separate conversation
Whether and how students may use these tools in their own work is a different question from how staff use them, and schools that try to answer both in one paragraph tend to answer neither.
It deserves its own discussion, with students in the room. A rule invented without them will be worked around by them, and everyone involved knows it.
Write it before you need it
The worst time to decide a school's position is during the incident that forces it — when a parent has complained, when a comment went out that should not have, when somebody is defending a decision they made alone with no guidance.
A page written calmly in a quiet week is worth a great deal more than a policy written in the middle of a complaint.
Where this leaves schools
The adoption has already happened. The question in front of most schools is not whether staff will use these tools but whether they will do so under a written position or without one — and only one of those is defensible when somebody eventually asks.
