Module 3 The Teacher’s Role in an AI Classroom
The module the rest of the programme stands on. Teachers work out the division of labour between themselves and AI — the teacher diagnoses, guides, challenges, contextualises, motivates and validates; AI analyses, suggests, personalises, practises, tracks and assists — and learn three tests for the decisions that cannot be delegated. They practise responding to AI suggestions with accept, adjust, reject or check first, start a decision log they keep for a week, and rehearse the two-minute explanation of AI in their classroom that students and parents will ask for.
By the end, teachers can…
- 1 State the teacher/AI verb split — teacher: diagnoses, guides, challenges, contextualises, motivates, validates; AI: analyses, suggests, personalises, practises, tracks, assists — and sort twenty classroom tasks onto it with a reason for each.
- 2 Explain “AI recommends, teachers decide” in their own subject’s terms, with one example from a class they actually teach.
- 3 Apply the three tests — context, consequence, relationship — to a given teaching decision and say which part AI can prepare and which part only the teacher can take.
- 4 Respond to an AI suggestion about a student with accept, adjust, reject or check first, and give the reason in one sentence that names what the teacher knew and the AI did not.
- 5 Explain to a Class 8 section, and separately to a parent at a PTM, in two minutes each, what AI is used for in the class and what it is not — and answer the common worries without promising results.
- 6 Keep a decision log for one week that records what AI suggested, what the teacher decided and why.
Session plan — 2 hours
| Time | Activity | Format |
|---|---|---|
| 0:00 10 min | Opening: what would you never hand over? Each teacher writes one teaching decision they would never hand to software and why. Pairs compare; the board collects them — “who sits where”, “whether a child is lazy or struggling”, “what to say to a parent”. That list is the spine of the module; the session returns to it at the end. | Discussion |
| 0:10 15 min | The verb split Teacher: diagnoses, guides, challenges, contextualises, motivates, validates. AI: analyses, suggests, personalises, practises, tracks, assists. One 9-B example per verb — the AI sorts forty Motion diagnostics by question; the teacher decides whether Meera’s graph errors are a concept gap or a bad week. The verbs are not ranks. They are a division of labour, and the order matters: AI first, teacher last. | Input |
| 0:25 20 min | Verb sort Groups get twenty task cards — “mark forty diagnostics by question”, “decide whether Meera’s drop is effort or home”, “choose which nine students get the fractions reteach”, “explain velocity–time graphs a third way”, “call a parent about attendance”, “set tomorrow’s homework”, “notice a student has gone quiet” — and place each on a board: teacher / AI / both in sequence. For “both”, they must say the order. Shareback on the cards that moved between columns. | Group work |
| 0:45 15 min | Where judgement cannot be delegated Three tests. Context: does the decision depend on things about the child a system cannot know? Consequence: does a wrong call cost the child — a label, a parent’s worry, a lost term? Relationship: does it depend on trust only the teacher holds? Any one of the three and the teacher decides. Three short cases: a “likely to fall behind” flag; an AI-drafted report-card remark; a recommended reteach list. And the honest counterpart: teacher judgement is fallible too — AI can challenge it with evidence; it cannot replace it. | Input |
| 1:00 25 min | Decision-log practice Each teacher receives six AI suggestions on the decision-log template (hypothetical AI suggestions from any tool — not a PrepGraph feature list): a flagged student; a recommended reteach list; an AI-drafted answer to a student’s doubt; a suggested question set for a diagnostic; a draft note to a parent; an AI-suggested mark for an answer. For each they write accept / adjust / reject / check first, a one-line reason, and what they would need to know. Pairs then compare — the same suggestion, two different decisions, both defensible. The disagreement is the point. | Hands-on |
| 1:25 25 min | Talking to students and parents Role-play in threes. First, a two-minute briefing to a Class 8 section: what we use AI for in this class, what it will not do, what I expect from you. Then a PTM: one teacher, one “parent” with a card — “Is the computer teaching my child now?”, “Will it replace you?”, “Is my child’s data safe?”, “Will it make her lazy?” — one observer with the checklist: honest, specific, no promises about marks. Rotate. Each teacher leaves with a one-page “how I explain AI in my class” script. | Group work |
| 1:50 10 min | Closing reflection Back to the opening board. Individually: the decision I will keep; the decision I will let AI prepare; the decision I will always check first. This becomes the first entry in the week-long decision log. | Reflection |
| Total 2 h · timings are a default; facilitators adapt to the group. | ||
Key ideas
AI recommends. Teachers decide.
Not a slogan — a procedure. Every AI output is a recommendation until a teacher acts on it; what turns it into a decision is the teacher’s judgement and the teacher’s name. The pattern-work comes first and is done by the machine. The knowing comes last and is done by you.
Three tests for what cannot be delegated
Context — does it depend on things about the child a system cannot know? Consequence — does a wrong call cost the child a label, a parent’s worry, a term? Relationship — does it depend on trust only you hold? If any one applies, the call is yours. Most decisions about a named child fail at least one.
Deciding is not agreeing
Accept, adjust, reject and check first are all decisions. Rubber-stamping a suggestion because it looks tidy is not deciding. Rejecting a good-looking suggestion with a reason — “I saw him run out of time” — is the professional skill this programme certifies.
Say it out loud
Students and parents fill silence with fear: the computer is grading me; the teacher is being replaced; my child’s marks are on the internet. A teacher who can explain in two minutes what AI is for in her class, and what it is not, keeps the trust that makes everything else possible.
Worked example
Nine names on a reteach list
After a ten-question diagnostic on Motion in 9-B, a results view — any tool — sorted by question shows the nine students who got both velocity–time graph questions wrong; read as a reteach list, that is the AI’s sorting done. Forty answer sheets sorted by question in seconds — the AI did what it is for.
The teacher reads the nine names. Seven she agrees with. Kabir is on the list, but she watched him run out of time on the last page — his graph errors are blanks, not mistakes. Sana is not on the list, but she has asked the same graph question twice in the doubts box this week and got both right on the diagnostic by luck.
Decision: seven kept, Kabir moved to “check first” (a two-minute chat tomorrow), Sana added. One line in the decision log: “Reteach list 9 → 9 (−Kabir, +Sana). Reason: timing / doubts pattern. Check Kabir Tue.” Total time: four minutes.
The AI did the sorting. The teacher did the knowing. Neither could have done the other’s job — and the child who benefited most was the one the list missed. That is the module in one example.
Assignment — counts towards certification
One-week decision log
For one week, keep a log of every time an AI output touched a teaching decision — any tool: a chatbot draft, a dashboard flag, an auto-drafted answer, a suggested list. For each entry record what the AI suggested; what you decided (accept / adjust / reject / check first); why, in one line; and what you knew that the AI did not. Finish the one-page “how I explain AI in my class” script from the role-play.
Deliverable: The decision log (at least five entries on the template) and the one-page script: a two-minute briefing for students and two-sentence answers to the three most common parent questions.
| Criterion | Meets the standard when… |
|---|---|
| Five real entries | Each entry is a real decision from the week, with the AI’s suggestion and the teacher’s decision both recorded — not invented scenarios. |
| Reasons that name what the teacher knew | Every entry has a one-line reason; at least one is an adjust or reject whose reason names something about the student or the class that the AI could not have known. |
| Delegation tests applied | At least two entries name which test — context, consequence or relationship — made it the teacher’s call rather than the tool’s. |
| An honest script | The script says what AI is used for in the class and what it is not, answers a parent worry directly, and makes no promise about marks or results. |
Where this lands in the PrepGraph workflow
Each certification task corresponds to a behaviour in the teacher portal — so the programme produces a working habit, not a slide deck. The pedagogy comes first; the surface implements it.
| Certification task | Pedagogical move | PrepGraph surface |
|---|---|---|
| Read the AI’s draft answer to a student’s doubt and decide: reply, correct, or take it to class | Accept / adjust / reject / check first on an AI suggestion | DoubtsStudents ask; the AI drafts an answer; you can reply, correct or take it to class. |
Materials
- •Twenty task cards for the verb sort (printable, facilitator pack)
- •Sorting board — teacher / AI / both in sequence — one per group, A3 or wall space
- •Six AI-suggestion cases for the decision-log practice
- •Decision-log template: suggestion / decision / reason / what I knew that the AI did not
- •Role-play cards: one Class 8 briefing brief, four parent questions, one observer checklist
- •Slides: the verb split; three tests for delegation
Facilitator notes
- •This is the module the rest of the programme stands on. If teachers leave with one sentence it must be “AI recommends, teachers decide” — said by them, in their words, about their own class. Ask for it in the closing reflection.
- •In the verb sort, most cards belong in “both in sequence” — AI first, teacher decides. Push groups to say the order out loud; a card marked “both” with no order has not been sorted.
- •In the decision-log practice, do not resolve disagreements between pairs. The same suggestion with two defensible decisions and two reasons is the lesson. Ask only “what did you know that made you decide that?”
- •In the role-play, ban promises. “It will improve marks” is off the table. The honest answer to “will it help?” is “it shows me sooner who needs what; what happens next is still my job.”
- •Do not demonstrate PrepGraph. The doubts row in the workflow table is for facilitators: it is where this habit will land in Module 8, when the AI drafts an answer and the teacher replies, corrects or takes it to class.
Bring the programme to your school
Tell us your board, grades and teacher count. We’ll reply within two working days with a cohort plan and the CPD record templates.