Why your 'AI policy' shouldn't be 14 pages long

I have read seventeen school AI policies in the past six months, written for everything from primary schools to selective entry secondaries. The longest was 22 pages. The most useful was one page. The relationship between length and usefulness was not what you'd hope.

The 22-page policy was thorough. It was also unread by the staff it was written for, because by the third paragraph it had quoted three pieces of legislation, two state-level guidance documents, and a UNESCO framework. It had been written, in good faith, by a person trying to anticipate every edge case. The result was a document that staff filed away and consulted approximately never.

The one-page policy was different. It said roughly four things. It said what staff were and were not allowed to use AI for, in plain language. It said what students were and were not allowed to use AI for, in plain language. It said what to do when something unexpected happened, with a person's name and a phone number. And it said when the policy would be reviewed next, by whom.

That was it. The whole thing fit on a single A4 page in 12-point font.

The reason it worked was not because it was simpler than the actual reality. The actual reality of AI in a school is genuinely messy and full of edge cases. The reason it worked was because the person writing it had made a deliberate choice about what level of abstraction belonged in the policy and what level of abstraction belonged in the conversations underneath it.

A 22-page policy attempts to encode every conversation in advance. It cannot succeed. The technology moves too fast. The classroom situations vary too much. The staff who are most likely to need the policy are the ones least likely to read 22 pages.

A one-page policy attempts something different. It establishes the principles, names the boundaries, and trusts the staff to handle the in-between situations using professional judgement, with named support when they need it. The principles are stable enough to last. The specifics get worked out in the staffroom, in faculty meetings, in moment-to-moment classroom decisions.

If you are writing or revising your school's AI policy this term, here is the test I would suggest before signing off on the final draft. Print it. Put it in front of a teacher who has been at your school for under a year. Ask them what they should do if a student submits a piece of writing that they suspect was AI-generated. If the teacher can give you a clear and confident answer in under thirty seconds based purely on what they have just read, the policy is doing its job.

If they have to flip pages, search for a section heading, or ask you a clarifying question, the policy is not yet doing its job, and adding more pages won't fix it.