LeasonAI logoLeasonAI
guides

Planning a Full Unit With AI in 30 Minutes (Without It Feeling AI-Made)

The five-stage AI-assisted unit planning workflow used by teachers who don't have Sunday afternoons anymore — with the exact prompts and where the human still has to do the work.

7/22/2026· 9 min read#unit planning#curriculum#workflows#prompts

Planning a Full Unit With AI in 30 Minutes (Without It Feeling AI-Made)

Most AI unit-planning tools produce something that looks like a unit and teaches like a wet paper towel. Generic objectives, mid activities, no thread between lessons. The problem isn't AI — it's the workflow. A good unit plan comes out of a specific sequence of teacher decisions, and AI can accelerate about 60% of that sequence if you feed it the decisions in the right order.

Here's the workflow, five stages, ~30 minutes end to end. It produces a unit skeleton that feels like yours because the decisions were yours.

Stage 1 — You decide the big idea (5 min, no AI)

Before you open any tool: write two sentences by hand.

AI is bad at these. They're small, specific to your school and students, and often connect to something you saw in a hallway last week. Skipping this step is why AI unit plans feel generic — they have no big idea, so nothing else can hang off one.

Stage 2 — AI drafts the arc (7 min)

Open a chatbot. Paste in your big idea and essential question, then ask:

I'm planning a [duration]-week unit for [grade + subject]. The enduring understanding is: [paste]. The essential question is: [paste]. Sketch a 3-phase arc for this unit: hook & noticing, building understanding, transfer & application. For each phase, name a 1-sentence goal, 2 possible activities, and one common misconception students bring in. Keep it to about 200 words total.

Read the output critically. The arc is usually about 70% right on the first pass. Cross out the activities that don't fit your students; keep the misconceptions (AI is good at these because they're well-documented).

Stage 3 — You choose the assessment (5 min, mostly no AI)

Before lessons, decide what students will produce that would prove they got the enduring understanding. Not a test — a performance. ("Design a hypothetical disturbance to a real ecosystem and predict whether it recovers, with reasoning.")

AI can help you brainstorm here — "give me five performance tasks that would show a student understands ecosystem resilience" is a fine prompt. But you pick which one, because you know what's doable with your resources, your schedule, and your specific students. Wrong pick here and the whole unit points nowhere.

Stage 4 — AI drafts the lesson sequence (8 min)

Now you have: big idea, essential question, arc, and summative task. Feed all four to the chatbot in one prompt:

Given: [enduring understanding], [essential question], [3-phase arc], [summative task]. Draft a [N]-lesson sequence. Each lesson: title, one learning goal, one hook, one main activity, one exit ticket idea. Use the misconceptions I listed as anchor points in specific lessons. Keep each lesson description under 60 words.

The reason this works better than "draft me a unit" is that it's constrained. The AI has your four decisions to build on, so it can't drift into genericness. What comes back will still need editing — usually 2 of the N lessons will feel off — but the sequence will be coherent because the anchors are.

Stage 5 — You do the read-through (5 min)

Read the sequence out loud. This step catches everything AI misses:

Rewrite the 2-3 spots that need it. Total elapsed time so far: about 25 minutes.

The last 5 minutes — the seams

Spend the final five minutes on the seams — the transitions between lessons, the connective thread. AI is bad at this because seams depend on what happened yesterday. Write a one-line "opener" and "closer" for each lesson, in your voice. This is what makes the unit feel like yours instead of a stitched-together document.

Why this beats the "generate my unit plan" tools

Those tools skip stages 1, 3, and 5 — the ones where teacher judgment is irreplaceable. What comes out is a shape without a spine. This workflow keeps the judgment steps for you and outsources the drafting to AI, which is what AI is good for.

The first time you run this, it takes about 45 minutes because you're learning the prompts. By the third unit, it's a real 30 minutes. Which is not a Sunday afternoon.