Professional judgement

The honest conversation about AI and lesson planning

Matt Dunn · May 25, 2026

At conferences, I talk to a lot of teachers. Most of them aren't scared of AI. What surprises me is how many are quietly using it and keeping that fact to themselves. They're generating lesson plans, drafting activities, differentiating texts, and telling nobody. Not their heads of department. Not their colleagues. Sometimes not even their partners.

That silence says something. It tells you the profession hasn't had an honest conversation about this yet. Teachers have figured out, on their own and without much support, that AI saves them real time. But they haven't figured out whether using it is something to be proud of or something to hide. So they do it quietly, cross their fingers, and hope nobody asks.

This article is an attempt at that honest conversation. Because the real risk isn't that teachers are using AI. It's that most of them are using it without a framework, and some of them are getting the results you'd expect.

The two layers every teacher needs to understand

Lesson planning has two layers, and getting them confused is where things go wrong.

The first layer is structural. You take a curriculum standard and turn it into a sequence with activities, resources, timings, and checks for understanding. This is real work. But it's largely mechanical. A competent teacher anywhere, teaching any class, would produce a broadly similar first draft. It's the scaffolding, and scaffolding follows rules.

The second layer is professional. It's knowing that the student in the third row will switch off the moment you introduce abstract text without a concrete hook first. It's knowing that the pair by the window had a falling out on Tuesday and can't work together this week. It's knowing the class got completely lost on a concept last term and needs a different entry point today. This layer is yours. It comes from being in the room, knowing your students, and reading what's happening in front of you. No AI can do it, because no AI has met your class.

AI doesn't create the problem of generic teaching. It just makes it faster and more visible.

Here's where the honest conversation gets uncomfortable. The profession tells itself a story about craft that doesn't always match what's actually happening in classrooms. A lot of lessons were generic long before AI arrived. And the teachers who resist AI loudest are sometimes the ones whose planning would benefit most from it, because the structural layer is eating all their time and leaving very little for the professional layer that actually matters.

Professional judgement isn't just about adaptation. It's also about knowing when not to use AI at all. There are lessons that need to come entirely from you, from your relationship with a class, from something that happened in the room last week. Knowing the difference is a professional decision, and it's one AI can't make for you.

A workflow that keeps you in the driving seat

The workflow that works has five steps. Each one keeps your judgement in the loop.

Unpack the curriculum with AI

Give it the standard or objective and ask it to surface the underlying concepts, common misconceptions, and the prerequisite knowledge students need. You're using AI as a thinking partner here, not a ghostwriter. The output should sharpen your understanding of what you're actually teaching, not hand you a plan to follow.

Ask for options, not a single answer

Ask AI for three possible sequences, each with a different emphasis: one concept-first, one skills-led, one inquiry-based. Then choose. The act of choosing is not a small thing. It's where your knowledge of your class, your curriculum context, and your own teaching strengths come in. Treat AI as a generator of options, not a source of answers.

Generate the lesson structure

Activities, timings, checks for understanding, resources. This is where AI saves the most time and where the output is most immediately usable. A specific, well-constructed prompt will give you a working draft in under a minute.

Adapt the draft to your class

Rewrite it through the lens of the students in front of you. Which activity won't work with your Year 4s on a Friday before lunch. Which text needs simplifying for your EAL learners. Which question needs reframing because you know exactly where this class will get stuck. This is where professional judgement earns its keep. Skip it, and you're not teaching your lesson. You're delivering someone else's.

Verify everything

Check the facts. Check the reading level. Check that the activity actually delivers the learning intention it claims to deliver. AI is confidently wrong often enough that verification isn't optional. It's part of the job now.

What this looks like in practice

Worked example

A Year 9 history teacher is planning a lesson on the causes of World War One. Without a framework, she types "write me a lesson plan on the causes of WW1 for Year 9" and gets back a starter, a reading task, and a discussion activity. It's fine. It's also a lesson any teacher anywhere could have been handed.

With the workflow, she starts by unpacking the topic. AI flags that students often confuse the long-term underlying causes with the immediate trigger of the assassination, a common misconception worth addressing directly. She asks for three sequence options, picks the one that builds from underlying tensions to the spark, and generates the lesson structure.

Then she adapts. Her class responded well to source comparison last term, so she swaps the generic discussion for that. Two students need a simplified version of the primary source, so she asks AI to produce one at a lower reading level. She verifies the dates and checks the attribution on the sources before she finalises anything.

Same starting point. Completely different lesson. The difference isn't AI. It's a teacher who stayed in the driving seat throughout.

Where leadership has gone missing

A word on the context most teachers are working in. School leadership is largely absent on this. Policies are vague or non-existent. Training is thin or hasn't happened at all. Most teachers are working AI out in snatched moments, without a framework, without guidance, and without anyone checking whether what they're doing is working.

That's not a teacher problem. It's a leadership problem. But it does mean you can't wait. The guidance isn't coming, or at least not quickly enough to be useful. You need your own workflow now.

The deskilling question

The worry that AI will deskill teachers is worth taking seriously. But it's pointing in the wrong direction.

Teachers who use AI well get sharper, not softer. The time saved on structural work goes into the part of teaching that actually matters: understanding your students and responding to them in real time. That's not deskilling. That's what teaching is supposed to look like.

Teachers who use AI badly do get deskilled. They stop questioning why a lesson is structured the way it is. They stop noticing when something is subtly wrong.

They start delivering plans rather than teaching classes. That's a real risk, and it's worth naming directly. It's also avoidable. The workflow is the difference.

Are you using AI to become a better teacher,
or just a faster one?