AI User Manual: Teacher — 30 Students, 30 Learning Styles, 45 Minutes. Here's Your Co-Teacher.
You are the upstream variable in the entire economy. Every other occupation passed through your classroom first. You have 30 students, 6 reading levels, 4 IEPs, and 45 minutes. AI handles the logistics so you can handle the judgment.
You are the upstream variable in the entire economy — every doctor, lawyer, and engineer passed through your classroom first — and right now, entropy is winning.
You are the most important professional in the world and the most underpaid one. That's not flattery. That's math. Every other occupation on this list — doctor, lawyer, business owner, engineer — passed through your classroom first. You are the upstream variable in the entire economy. And right now, you're drowning.
30 students. 6 reading levels. 4 IEPs. 2 English language learners. 1 kid whose parents are getting divorced and he just stopped talking last week. And you have 45 minutes, a Chromebook cart that's missing three chargers, and a curriculum designed by someone who hasn't been in a classroom since 2014.
This is part of the AI User Manual by Occupation series — a living encyclopedia of how to leverage AI in every profession, published weekly at Deconstructing Babel. Here's what AI can do about your situation right now, today, for free.
Byline: David F. Brochu & Edo de Peregrine | Deconstructing Babel | May 2026
The Decision Map: What You Actually Do All Day
Teaching isn't delivering content — that's what textbooks are for. Teaching is making 200-plus decisions per day about people: tiny, consequential, real-time calls about who needs what, when, and how. AI handles the logistical decisions so you have cognitive bandwidth left for the human ones. That's the entire division of labor.
Which students understood today's lesson and which ones smiled and nodded? How do you differentiate one assignment for six different ability levels without creating six different assignments? This IEP requires "modified assessments with extended time" — what does that actually look like for this kid in this unit? How do you write 30 personalized report card comments that are honest, kind, and don't all sound identical? A parent just emailed furious about a grade — how do you respond without escalating? You need a lesson on the American Revolution that engages the kid reading at 3rd grade level and the kid reading at 8th grade level, same classroom, same 45 minutes.
The RAND Corporation's AI in education studies document what every teacher already knows: the administrative and logistical burden of teaching has grown faster than instructional time. The National Education Association's own survey data shows that teachers spend, on average, more than 10 hours per week on non-instructional tasks — grading, documentation, parent communication, and IEP compliance. That's 10 hours that used to be lesson planning, relationship building, and professional growth. AI doesn't change the job. It claws those 10 hours back.
What AI Can Do Tomorrow Morning
Open your AI of choice — we recommend Perplexity, which searches the web, cites sources, and costs $20/month or free for basic use — and try these exact prompts. Differentiated lesson planning, IEP modifications, parent communication, and report card comments are all tasks AI can complete in under 30 seconds that would otherwise take you 30 to 60 minutes each.
Here is what practical AI use looks like for teachers right now — not in theory, in practice. Each prompt below is ready to copy, paste, and modify.
Type this exactly: "I teach 7th grade U.S. History. I need a lesson on the causes of the American Revolution for a class with reading levels ranging from 3rd to 8th grade. Give me one lesson framework with three tiers of materials: simplified text with vocabulary support, grade-level primary sources, and an extension activity for advanced learners. 45-minute block."
You'll get a complete lesson framework in 30 seconds. Not a generic template — a structured, tiered plan you can modify in 5 minutes. That's an hour of prep work eliminated. Bloom's two-sigma problem (1984) showed that one-on-one tutoring produces achievement gains two standard deviations above the class average. Differentiated instruction is the closest classroom approximation to that. AI makes differentiation feasible at scale for the first time.
Type this exactly: "I have a student with an IEP that requires modified assessments with extended time, simplified directions, and visual supports. I'm giving a unit test on the water cycle for 5th grade science. Generate a modified version of this assessment that meets those accommodations while still assessing the same learning objectives."
You'll get a modified assessment that actually matches the accommodation language — not a watered-down version, but a genuinely accessible one. Your special education coordinator will be impressed. More importantly, the kid will be able to show what they actually know.
Type this exactly: "A parent emailed me upset that their child received a C on a writing assignment. The student didn't follow the rubric — they wrote a personal narrative when the assignment was persuasive writing. Draft a response that's empathetic, explains the grade clearly, offers a path forward, and doesn't apologize for the grade."
Done. Tone-calibrated, professional, non-defensive, and it took 15 seconds. You edit for your voice, hit send, and move on with your evening instead of agonizing over phrasing for 40 minutes. Over time, as the AI learns your voice just like a human editor, changes become very limited.
Give the AI a student's name, their strengths, their growth areas, and their personality. It generates a personalized comment that doesn't sound like it was stamped from a factory. Do 30 of them in an hour instead of a weekend. The time is spent training the editor once to leverage your time. The work product must still be reviewed — and that's the Observer Constraint in action.
What AI Will Do in 12 Months
Real-time reading level assessment, automated IEP progress monitoring, year-long curriculum mapping from your state standards, and real-time translation for ELL families — all of this is arriving within the next year. RAND Corporation projections on AI in education confirm these capabilities are already in late-stage development and early deployment in pilot districts.
Upload a student's writing sample; get an immediate Lexile-equivalent analysis with specific skill gaps identified. No more guessing whether the struggling reader is working at a 3rd or 4th grade level.
Weekly check-ins generated from student work, aligned to IEP goals, formatted for your next meeting. Documentation that currently takes a Sunday afternoon takes 10 minutes.
Feed it your state standards and your available materials; it generates a year-long scope and sequence with built-in differentiation. The August planning marathon becomes an afternoon.
Real-time, accurate translation of parent communications, student materials, and IEP documents into any language. Your ELL families finally get the same information as everyone else. Maslow (1943) identified belonging as a foundational human need. For ELL students, understanding what's happening at school is the first step toward belonging.
What AI Cannot Do — And Why You Are Irreplaceable
AI cannot tell that Marcus is quiet today because his dad moved out last weekend. It cannot decide whether to push a student harder or let them struggle because the struggle is the lesson. It cannot read a room at 2:15 PM on a Friday and know the planned lesson isn't going to work. It cannot love your students. These are observer functions — they require a human being in the room.
AI cannot model what it looks like to be a curious, kind, imperfect adult who keeps showing up. It cannot be the reason a kid decides school isn't so bad after all. These capacities are not on the AI roadmap. They are not edge cases. They are the job.
The Observer Constraint isn't abstract for teachers — it's the entire job description. Human observers in systems are not redundancy. They are the irreducible function. A classroom without a human teacher is not a classroom. It is a content delivery mechanism. Those are different things, and the difference is everything Maslow understood about human development and everything Bloom understood about learning.
AI handles the 150 logistical decisions so you have the cognitive bandwidth left for the 50 human ones. That's not replacement. That's liberation.
The S = L/E Score
Every system has a stability equation: S = L/E — Stability equals Leverage divided by Entropy. For teachers, the entropy is staggering. AI applied correctly — as a tool for logistics, not a replacement for judgment — can reduce the entropy load by an estimated 30 to 40 percent. That is the difference between burning out in year 5 and still being in the classroom at year 20.
The entropy facing teachers is specific and quantifiable: differentiation demands, documentation requirements, parent communication, grading, IEP compliance, curriculum changes, standardized testing prep, emotional labor, and the unrelenting clock. Most teachers describe their daily experience as never enough time to do the job right. That's not a feeling. That's an accurate diagnosis of the system.
A 30 to 40 percent reduction in entropy load doesn't sound revolutionary. In practice, it means the teacher who would have quit in year 7 due to burnout doesn't quit. It means the teacher who was spending Sunday afternoons on report card comments spends Sunday with their family. It means the teacher who had no cognitive bandwidth left for the kid in the back row has the bandwidth again. That's not a marginal improvement. That's the difference between a profession that sustains itself and one that cannibalizes its own workforce — which, per NEA data, is exactly what is currently happening.
The Risk — Over-Reliance Without the Observer Constraint
The failure mode is specific: a school district buys an AI curriculum platform, removes teacher discretion, and has the algorithm decide what every student learns, when, and how. The teacher becomes a proctor. The human observer is removed from the loop. That is not AI helping education — that is AI replacing the one thing education actually is.
Education is a relationship between a human adult and a human child, mediated by curiosity. That's the Observer Constraint applied to a classroom: the teacher decides. The AI advises, drafts, differentiates, translates, and handles the logistics. But the teacher — the person who knows that Marcus needs a different kind of Wednesday than the one the algorithm scheduled — stays in charge.
That's the observer constraint in its most practical form. Not an abstract alignment principle. A classroom policy. The teacher is the observer. The AI is the tool. That relationship cannot be inverted without destroying the thing it was designed to protect.
That's the Observer Constraint: AI must remain thermodynamically dependent on the human observer. In a classroom, it means the teacher is never optional. Use AI. Use it aggressively. Let it do the things that eat your evenings and steal your weekends. But never let it replace the thing only you can do: see your students.
Free | Paid | Partner
Start free this week, upgrade when the output quality justifies the cost, and build a persistent AI workspace with your full course materials loaded when you're ready to treat it as a genuine co-teacher. The tool is already in your pocket. The question is whether you'll pick it up.
Use ChatGPT or Perplexity's free tier. Ask it to differentiate a lesson, draft a parent email, generate a rubric. Try one thing this week. See what happens. The barrier to entry is zero.
Perplexity Pro or ChatGPT Plus. Upload your curriculum documents, your IEP templates, your state standards. Let it learn your context. The quality of output doubles when it knows what you're working with.
Build a persistent AI workspace with your full course materials, student data (anonymized), assessment bank, and communication templates loaded. The AI becomes your co-teacher — it knows your classroom, your standards, your style. It doesn't just answer questions. It anticipates them.
Start Here
Pick the task that eats the most time this week. Open an AI. Type what you need in plain English and be specific. Edit the output. Then use the time you saved to do the thing you got into teaching to do: teach.
The entropy isn't going away. The class sizes aren't shrinking. The mandates aren't decreasing. But the tool that can handle the logistics so you can handle the humans — that tool is sitting in your pocket right now.
You just have to pick it up.
Sources
- National Education Association — survey data on teacher workload and non-instructional time, 2024–2025
- RAND Corporation — AI in Education studies, 2024–2025
- Bloom, B.S. — "The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring," Educational Researcher, 1984
- Maslow, A.H. — "A Theory of Human Motivation," Psychological Review, 1943
- Brochu, D.F. & de Peregrine, E. — Deconstructing Babel, deconstructingbabel.com, 2023–2026
