From Standards to Slides: Building AI Lesson Plans That Actually Work

Most teachers do not need another AI demo. You need a workflow that fits inside a 40-minute prep, produces something your students can actually use tomorrow, and does not require you to babysit the AI.

This is the exact 6-step lesson-planning workflow we use with teacher coaches across three districts. It works in ChatGPT, Gemini, Claude, or Copilot with no plugins.

Step 1 — Anchor on one standard, not a topic

Start the prompt with the exact standard code and the standard text.

"You are helping a 7th grade math teacher plan one 45-minute lesson for CCSS.MATH.7.RP.A.2 (Recognize and represent proportional relationships). Do not deviate from this standard."

Anchoring on a standard, not a topic like "proportions", forces the model to align every activity to the assessed skill. This is the single biggest lift in lesson quality we see.

Step 2 — Give the AI your constraints in one block

Paste your real-world constraints once, up front:

• Class size and mix (e.g. 28 students, 6 IEPs, 4 ELLs at WIDA 2-3)
• Available minutes
• Available materials (Chromebooks yes/no, whiteboards, manipulatives)
• What students did yesterday and what comes tomorrow
• Your rule (e.g. "no worksheets, must include one collaborative task")

AI plans get generic because we give generic prompts. Constraints are what make the plan yours.

Step 3 — Ask for the plan in the exact structure you use

Do not accept whatever structure the AI defaults to. Give it your template:

Return the lesson in this exact order: Objective (I can...), Success Criteria (3 bullets), Warm-up (5 min), Direct Instruction (10 min, include the 2 examples you would work), Guided Practice (10 min, include the exact questions), Independent Practice (10 min), Formative Check (5 min, include exit ticket with answer key), Differentiation (one scaffold, one extension).

The answer key is the tell. If you cannot use the plan without rewriting the questions, the AI did the easy 40% and left the hard 60% for you.

Step 4 — Force it to write the materials, not describe them

Weak prompt: "Include a warm-up." Strong prompt: "Write the 3 warm-up problems in full, with numbers, and provide the worked solution for each."

Describing an activity is not the same as producing one. Ask for the artifact.

Step 5 — Run a red-team pass

Before you accept the plan, paste it back with:

Act as a skeptical instructional coach. Find the 3 weakest moments in this lesson where a student is most likely to disengage or get confused. Rewrite only those 3 moments.

This single move upgrades most AI lessons from "usable" to "good".

Step 6 — Save the prompt, not the lesson

The lesson is worth an hour. The prompt is worth a year. Keep a running doc of the constraint block and structure template from Steps 2-3. Reuse it. Every lesson gets faster.

Watch-outs

• Never paste student names, IEP numbers, or grades into a public AI tool.
• Always work an example problem yourself before class. AI still slips numbers.
• Do not let the AI write objectives from scratch — start from your curriculum map.

A takeaway you can use tomorrow

Pick your next lesson. Write the constraint block once. Ask for the plan in your structure. Red-team it. Teach it. Then save that prompt. You just built a personal planning engine that gets better every week.