Here is a sentence I have heard at every leadership conference for the last eighteen months. We need to figure out how to police student use of AI. I want to push back on that sentence, hard, because I think it has done more damage to AI conversations in international schools than any other single phrase.
When you frame the teacher as a police officer, three things happen. The first is that the teacher's relationship to the technology gets adversarial. The second is that the assessment redesign conversation becomes about detection rather than purpose. And the third — the worst one — is that students learn to think of AI as something they do behind their teacher's back rather than alongside them. None of those outcomes are what any of us want. All three follow directly from the framing.
Let me describe what teaching with, not against, actually looks like in a classroom.
A Year 9 Biology teacher I worked with last term redesigned her photosynthesis unit. The old unit had a take-home essay assessment. She knew, and her students knew, that the take-home essay had stopped being a meaningful signal of understanding. So she rebuilt the unit around three things. Lessons used AI. Homework explicitly used AI. Assessment was in-class, oral, with the AI in front of the student, and she asked them to explain the diagrams they'd generated. Her students did better on the unit test than the previous year's cohort. Her parents understood what their children were doing. And the teacher spent less time grading take-home essays she didn't trust.
That is what teaching with, not against, looks like. It is not permissive. It is not policing. It is a deliberate redesign of what we ask students to demonstrate, given what they actually have access to.
Differentiation is the same story. A teacher who policies AI cannot differentiate at scale, because differentiating means producing variant materials for variant learners and there is no time in any teacher's week for that. A teacher who teaches with AI generates three versions of every starter activity in twenty minutes — one for the student who needs scaffolding, one for the student who's on track, one for the student who needs the extension. The struggling student gets a version pitched at their actual reading level. The advanced student gets a version that pushes them to a question their teacher would not have had time to write. The teacher walks into the classroom with three lessons instead of one, and not one of those students has been short-changed.
Feedback is the same story again. A teacher who policies AI gives feedback the way teachers have always given feedback — in the margin, days later, on a draft the student has already mentally moved on from. A teacher who teaches with AI sets up an in-class loop where students get a draft, get AI feedback on it in real time, take that feedback to their teacher for the conversation that actually matters, and revise. The teacher's role shifts. They are no longer the source of all feedback. They are the curator of which feedback is worth listening to. That role is, if anything, more important than the previous one. It is also more interesting to do for thirty years.
I want to close with the principle that organises all of this. Make the teacher the hero, not the gatekeeper. The hero teaches their students how to use a powerful tool well. The gatekeeper tries, and fails, to keep a powerful tool out of the room. We have known for a long time which of those two stances scales. We have just been slow to apply that knowledge to AI.
If you are designing a policy or a programme right now, ask yourself which version of your teachers it produces. The hero version produces classrooms your students will remember twenty years from now. The gatekeeper version produces classrooms they spent four years learning to work around. Pick the first one.
Thank you.