Plan a Whole Unit in One Sitting: A 4-Step AI Workflow for Medium-Term Planning

Most AI advice for teachers stops at the single lesson. That's useful, but it's not where the hours go. The real grind is medium-term planning — sitting down before a new topic and mapping six or eight weeks of learning so it actually builds, hangs together, and lands where you need it to.

Done from a blank page, that's an evening gone. Done as a workflow, it's about half an hour. The trick is to feed AI one step at a time instead of asking for "a unit plan" all at once — that always gives you generic mush.

Here's the four-step sequence I use. Run it in ChatGPT or Claude in a single chat so each answer builds on the last.

Step 1: Set the destination first

Don't start with activities. Start with what students need to know and be able to do by the end. Give the AI your outcomes and let it turn them into a learning sequence.

You are helping a New Zealand [subject] teacher plan a unit for Year [level]. The unit is [topic] and runs for [X] weeks, [Y] lessons a week. Here are the key outcomes students must reach by the end: [paste your curriculum outcomes or achievement objectives]. Break these into a logical teaching sequence — what has to be understood first, what builds on what. Give me the sequence as a numbered list of learning steps, not activities yet.

This gives you the spine of the unit. Read it critically — you know your class, the AI doesn't. Reorder anything that's out of sequence before moving on.

Step 2: Turn the sequence into a week-by-week map

Good. Now map those learning steps across the [X] weeks. For each week give me: the focus, the key concept or skill, and one sentence on what success looks like by Friday. Keep it realistic for [Y] lessons a week — don't overload it.

That "don't overload it" line matters. Left alone, AI crams three weeks of content into one. Naming the constraint fixes it.

Step 3: Build in the checkpoints

A plan without assessment is just a wish list. Ask for the checks before you fill in activities, so the teaching points toward them.

Now add formative checkpoints. For each week, suggest one quick, low-effort way I can check whether students have hit that week's success criteria — something that takes under 10 minutes and doesn't create a marking pile. Flag the single most likely misconception for each week too.

The misconception flags alone are worth the prompt — they tell you where to slow down before you've taught a single lesson.

Step 4: Generate the lesson skeletons

Only now do you ask for activities — and because the AI already has the whole map, they'll actually connect.

For Week [1], give me a lesson-by-lesson skeleton: a hook, the main learning activity, and how it links to next lesson. Keep resources low-prep — I don't want to build 10 new worksheets.

Run that per week as you need each one. You don't have to generate all eight weeks today; the map from Step 2 means you can come back and pull the next skeleton in two minutes whenever you're ready.

One habit that makes it stick

Paste the final map into a Google Gem or a saved project so it's there every time you open that unit. When a lesson flops or a week runs long, tell the AI what happened and ask it to adjust the rest of the map. A living plan beats a perfect one that's already out of date.

Start with one unit — the next topic you're dreading. Have a go, what's the worst that happens? You'll have a sequenced, checkpointed plan before your coffee's cold, and the rest of the term gets easier from there.

TeachSmarterAI helps New Zealand teachers use AI tools effectively. New posts every week.