Three readiness levels, one source standard, one prompt. By the end of this 45-minute lesson, you will leave with a real plan for tomorrow's class — and a workflow you can repeat for every unit.
Take one state standard and produce a single-day lesson plan with three differentiated tiers (below-grade, on-grade, above-grade) in under ten minutes — without putting any student-identifying information into a general-purpose AI tool.
If you teach four sections a day and you write one lesson plan per section per night, you write twenty plans a week. Most plans share the same scaffolding: an objective, a hook, direct instruction, guided practice, independent practice, and an exit ticket. The structure is identical. What changes is the standard, the content, and the readiness of the students.
That structural sameness is exactly what AI is good at. A well-instructed model can generate the scaffolding for a lesson plan in seconds. Your job shifts from writing the plan to directing the plan and editing the parts that require your professional judgment about your students. That is a much smaller cognitive load and a much shorter evening.
The trap is the generic output. If you ask an AI tool "write me a lesson plan on photosynthesis for 7th grade," you will get a lesson plan, but it will read like every other AI lesson plan ever generated: a competent, soulless, off-the-shelf draft that does not feel like yours. The fix is not to give up on AI — it is to give the model the specific structural information it needs to produce something that is actually useful in your classroom.
Every effective lesson-planning prompt has three blocks: the standard (what students must learn), the context (what your classroom looks like), and the output spec (what shape the response must take). Skip any one of those and the output regresses to generic. Include all three and the output is usable on the first pass roughly 80% of the time.
Here is the template I use, scrubbed of any student information. Copy it, fill in the brackets, and paste it into your AI tool of choice.
You are a curriculum designer helping a [GRADE LEVEL] [SUBJECT] teacher
plan a 45-minute lesson.
STANDARD:
[Paste the exact text of one state standard or learning objective.]
CLASSROOM CONTEXT:
- Class size: ~28 students
- Approximately 1/3 of students read below grade level
- Approximately 1/4 of students are English Language Learners
- Approximately 4 students have IEPs with reading-comprehension goals
- The lesson follows [PRIOR LESSON] and precedes [NEXT LESSON]
OUTPUT REQUIREMENTS:
Produce a lesson plan with these exact sections:
1. Learning objective (one sentence, student-facing)
2. Hook (5 minutes) - a concrete scenario or question
3. Direct instruction (10 minutes) - bullet points only
4. Guided practice - TIERED into three versions:
a. Below grade level: scaffolded with sentence frames
b. On grade level: standard task
c. Above grade level: extension with one open-ended question
5. Independent practice (10 minutes)
6. Exit ticket - three questions, increasing difficulty
7. Materials list
Constraints:
- No worksheets longer than one page
- All vocabulary defined in plain language at first use
- No student names, no real student data
Run that prompt with one standard. The output will be roughly 70% usable. Your editing pass — adding the specific examples your students will recognize, swapping in the vocabulary you have been building all year, adjusting the timing — takes about three to five minutes. Total time: under ten.
This is the part of the lesson that I will repeat in every day of this course because it matters. The Family Educational Rights and Privacy Act (20 U.S.C. § 1232g) restricts disclosure of "personally identifiable information" from a student's education record. That includes — but is not limited to — student names, parent names, addresses, dates of birth, IEP content, behavioral records, and grades tied to an identifiable student.
Most general-purpose consumer AI tools (the free tier of ChatGPT, Claude, Gemini, Copilot) train on user input by default unless explicitly disabled, and even with training disabled, the conversation transcript is processed and stored on third-party servers that may not be under a FERPA-compliant data-processing agreement with your district. Pasting student-identifying content into a consumer chatbot can be, in many districts, a FERPA violation regardless of whether anything bad ever happens.
The rule of thumb that keeps you safe: the lesson plan you draft today contains standards, tasks, and generic class composition ("about 1/3 of my class reads below grade level"). It contains no names, no IEP content, no behavior notes, no specific student work, and no test scores. If you want to paste an essay for feedback, strip the name first. We cover the safe pattern for that on Day 2.
Pick one standard you will teach in the next two weeks. Use the three-block prompt template above to generate a tiered lesson plan. Edit it for five minutes. Then teach it. Notice exactly which parts you kept, which parts you rewrote, and which parts you cut. The pattern of your edits is the most valuable diagnostic in this entire course — those edits are the part of teaching the model cannot do, and they tell you where your professional judgment is most concentrated.
Bring that lesson plan with you to Day 2, where we apply the same workflow to feedback on student writing.