How to Run an AI INSET Day Your Staff Will Actually Thank You For

Most INSET days are tolerated. A good AI INSET day is the rare one staff talk about positively afterwards — because it's one of the few sessions that can hand them time back rather than take it. But that only happens if you structure it right. Here's how to run one that actually lands, and the mistakes that turn it into another wasted training day.

Decide what the day is for — and it isn't "awareness"

The biggest INSET mistake is aiming for awareness. A day of slides about what AI is, ethical considerations, and future possibilities leaves staff exactly where they started, minus a day. Set a harder goal: every member of staff leaves having produced something real they'll use next week. A drafted parent letter. Three differentiated versions of an activity. A bank of report comment starters. If the day ends with artefacts, it worked. If it ends with opinions, it didn't.

A structure that works

• First hour — the rules, briefly. Cover the non-negotiables: no pupil data in any tool, human review of everything, which tools your school has approved. Keep it to an hour; the rules matter enormously but they don't need a morning. Have your one-page staff guidance printed and on every table.
• Rest of the morning — hands-on by role. Split by job, not by department. Teachers work on planning and resources, TAs on differentiation and interventions, office staff on letters and comms, SLT on documents. Everyone works on their own real tasks — this week's actual letter, next week's actual lesson — not invented exercises.
• Afternoon — produce and share. Each group builds something usable and shows it back. The sharing matters: a Year 2 teacher seeing what the office produced sparks ideas nothing on a slide can.
• Last 30 minutes — commitments, not evaluations. Skip the feedback form. Instead: each person names the one task they'll use AI for in the next fortnight. Write them down. That list is your follow-up plan.

The mistakes that waste the day

• Leading with fear. Open with an hour of risks and horror stories and you'll spend the rest of the day dragging staff back from "this is dangerous, I won't touch it." Cover the rules clearly and briefly, then move to the useful part.
• Demonstrating instead of doing. Watching someone else use AI is nearly worthless. The value is in staff typing their own prompts, hitting their own problems, and fixing them with support in the room.
• Tool overload. One tool, used well, for the whole day. Five tools in one day means none of them stick.
• No follow-up plan. This is the big one — more on it below.

Who should run it?

Be honest about your in-house capacity. If a member of staff genuinely knows this ground well — not just "uses ChatGPT a bit," but understands the data rules and can troubleshoot live — an internal session can work well. If not, bring someone in. A confident external lead who's done this in other schools will get more from the day than a nervous volunteer working one page ahead of the room, and the cost of a facilitator is trivial against the value of a whole staff day. What matters either way: whoever leads it must run a hands-on session, not a presentation.

The follow-up is worth more than the day

Here's the truth most schools learn the expensive way: a brilliant INSET with no follow-up produces almost nothing by half-term. Enthusiasm decays fast, old habits are comfortable, and the first time someone hits a snag with no support, they quietly stop. The INSET is the launch, not the programme. Plan from day one for what happens in weeks two, four and eight: a staff-meeting check-in against those written commitments, a shared place to collect prompts that worked, and a named person people can actually ask. Ten minutes of structured follow-up a fortnight protects the entire investment of the day. Our guide to introducing AI in your primary school covers the longer arc this fits into.