The programme

Six sessions, three strands, one written policy

Each session below is described the way we describe it to staff: what happens, what you bring, and what you walk out holding. Nothing here is a demonstration of a product. The tools change every term; the judgement does not.

Session 1 · Strand one

75 minutes · whole staff

What it actually is, in ten minutes

Enough to reason about it, no more. Staff learn what a large language model is doing when it writes, in terms a non-specialist can repeat to a class: prediction, not knowledge. Then the three things every teacher needs to hold in mind: what it is good at, what it fabricates, and why it sounds confident when it is wrong. We demonstrate a confident, fluent, entirely invented fact and let the room catch it.

Staff bring: Nothing. This session is deliberately low-friction.
Staff leave with: A shared vocabulary for the rest of the programme, and one fabricated answer they caught themselves, which does more than any slide.

Session 2 · Strand one

90 minutes · whole staff, then departments

Planning and marking support

Where it saves a teacher real hours, and where handing over is a mistake. We work through schemes of work, differentiation for a mixed-attainment class, and first drafts of written feedback. The rule staff leave with: it drafts, you decide. A draft scheme of work is a time saving; a draft judgement about a child is a safeguarding problem. Teachers try both in the session, on their own subject, and mark the difference.

Staff bring: One unit they are about to plan, and one class they know well.
Staff leave with: A drafted scheme-of-work skeleton for that unit, and a department-agreed line between acceptable drafting and unacceptable delegation.

Session 3 · Strand two

Half day · departments, the core session

Rewriting an assessment

Teachers bring three of their own tasks and leave with all three rewritten so they still measure learning. The signature method: put the AI’s answer into the task. Students mark it, find its errors, argue with it. The redesigned task requires exactly the knowledge the original was meant to test, and a chatbot can no longer do the work for the student, because the chatbot’s work is the material under examination.

Before

Describe how changing the concentration of hydrochloric acid affects the rate of reaction with magnesium ribbon.

Answered in one prompt. No knowledge required.

After

An AI tool answered the question above. Its answer states that doubling the concentration always doubles the rate, and describes bubbles of oxygen. Mark the answer: correct every scientific error, explain the collision theory it should have used, and write the conclusion the evidence would actually support.

Requires exactly the knowledge the original was testing.

Before

How does Shakespeare present Macbeth’s ambition in Act 1?

Answered in one prompt. No knowledge required.

After

An AI tool answered the question above. Check every quotation it uses against the text: two are fabricated and one is pulled from the wrong scene. Then argue with its reading: write a paragraph showing where its account of Lady Macbeth’s influence is too simple, using quotations you have verified yourself.

Requires exactly the knowledge the original was testing.

Staff bring: Three tasks they actually set: homework, classwork or an assessment. Shared through the school’s own systems before the session, never through a public form, and never with student names or work attached.
Staff leave with: All three tasks rewritten and ready to use, plus the method, so the fourth and the fortieth take twenty minutes.

Session 4 · Strand two

75 minutes · whole staff

Detection, and why it doesn’t work

What detectors actually measure, their false-positive rate, and why accusing a student on a detector score is a disciplinary risk the school will lose. We run real texts through real detectors in the room, including work we know to be human, and watch them fail. Then the constructive half: the defensible process a school should follow instead, built on conversation, drafting history and task design rather than a percentage.

Staff bring: Any suspicion stories they are willing to share anonymously. Every staffroom has them.
Staff leave with: A step-by-step process for suspected misuse that a head can defend to a parent, a governor or a panel.

Session 5 · Strand three

90 minutes · whole staff, with safeguarding lead

Data and safeguarding

Age limits and terms of service, read properly for once. What must never be entered into a consumer chatbot: student names, safeguarding notes, medical information, unpublished results. How the tools store and reuse what is typed into them. And the harder conversation: how to talk to students about what they disclose to a system that remembers everything and answers like a friend.

Staff bring: Questions from their own practice. The safeguarding lead attends, and their policies frame the discussion.
Staff leave with: A clear never-list, a shared script for talking to students, and the safeguarding lead’s amendments noted for session six.

Session 6 · Strand three

Half day · staff draft, leadership chairs

Writing your policy

The staff draft the school’s own AI policy in the room. We bring the structure and the difficult questions; the school brings its voice, its context and its red lines. By the end there is a working draft covering classroom use, assessment, data, and the disciplinary process from session four, plus the parent letter and the student-facing one-pager in skeleton form. A policy the teachers wrote is a policy the teachers follow.

Staff bring: Everything from sessions one to five, and their honest objections.
Staff leave with: A drafted policy, a drafted parent letter, and a named owner and date for each remaining document. We edit and return the final versions within ten working days.

After session six

The follow-up is where the real questions arrive

A term later, staff have actually used the rewritten tasks, fielded the parent queries and hit the cases nobody predicted. The follow-up session, included in the annual contract, works through those live cases and patches the policy where reality found the gaps.

Ready when you are

Book a call

Tell us about the school and we will tell you honestly whether the programme fits, and in which format.

Book a call