Playbook: Teachers and Educators

Teachers lose most of their evenings to three things: preparing material, writing feedback, and adapting the same lesson for students at very different levels. AI is genuinely good at all three — as long as the teacher stays the one who decides what's true, what's fair and what grade something deserves.

---

1. Where AI helps, and where it mustn't lead

Use AI for	Keep human
First drafts of lesson plans and activities	Final judgement on what fits your class
Differentiated versions of the same material	Knowing which student needs which version
Feedback drafts against your rubric	The final grade
Practice questions and worked examples	Checking every answer key
Parent communication drafts	Anything sensitive about a specific child

Privacy rule: don't paste student names, marks, or personal circumstances into consumer AI tools. Use initials or "Student A", and your school's approved tools for anything identifiable. Children's data gets extra protection under India's DPDP Act.

2. The 60-second lesson plan

Create a 45-minute lesson plan for [grade] students on [topic],
aligned to [board: CBSE / ICSE / state board / IB].

Include:
- Learning objectives (2-3, measurable)
- A 5-minute hook connected to students' everyday life in India
- 3 activities (one individual, one pair, one whole class)
- Materials needed (assume a normal classroom, no devices)
- A 5-question exit quiz with answers
- One adaptation for students who finish early, one for students
  who are struggling

Flag any fact I should verify against the textbook.

Then check it against your textbook. AI mixes up curriculum details between boards and years.

3. Personalised learning: one lesson, three levels

Differentiation used to take hours. Now:

Here is a reading passage/explanation for [topic]:
[PASTE]

Rewrite it in three versions:
1. Support: simpler vocabulary, shorter sentences, a glossary of
   5 key words, same core concepts
2. Core: as original, with 3 comprehension questions
3. Extension: adds one deeper idea and 2 higher-order questions

Do not change any facts. Keep all three recognisably the same
lesson so the class can discuss it together.

Also useful: the five-level explanation — ask for the same concept explained for a 5-year-old, a teenager, a college student and an expert. It shows you where the real conceptual jump is.

4. Feedback at scale — glows and grows

You are helping a teacher draft feedback. The teacher will review
and edit before sharing.

Rubric: [PASTE RUBRIC]
Student work (anonymised): [PASTE]

Write:
- 3 "glows": specific things done well, quoting the work
- 1 "grow": the single most important next step, with an example
  of what better looks like
- Tone: warm, direct, age-appropriate for [grade]
- Under 120 words

Do NOT suggest a grade.

Never let AI assign the final grade. It can't see effort, context or growth over time, and models can be inconsistent or biased — for instance marking non-standard English more harshly.

5. A Socratic tutor that doesn't give answers

Set up a custom assistant (Custom GPT, Claude Project or Gemini Gem) with these instructions:

You are a Socratic tutor for [grade] students learning [subject].
- Never give the final answer, even if asked repeatedly.
- Ask one question at a time. Wait for the student's reply.
- If the student is stuck twice, give a small hint, not the answer.
- When they get it right, ask them to explain why in their own words.
- If a question is outside [subject] or asks for personal advice,
  tell them to ask their teacher.

Test it yourself by trying to trick it into giving the answer before students use it. Check your school's policy on students using AI tools, and their age limits.

6. Designing assignments for an AI world

Trying to catch AI use with detectors doesn't work — they produce false accusations, particularly against students writing in a second language (see the safety module). Redesign instead:

• AI-assisted by design: students use AI to generate an outline or counter-arguments, then write and defend their work in class.
• Process over product: collect drafts, notes and a short reflection on what they changed.
• Personal and local: assignments grounded in their own experience, community or class discussion.
• Oral checks: a two-minute conversation about their work reveals understanding quickly.

This teaches what students will actually need: using AI well, and thinking critically about what it produces.

---

⚠️ Common mistakes

• Letting AI give final grades. Draft feedback, yes; judgement, no.
• Using student names and details in unapproved tools.
• Trusting curriculum facts without checking the textbook.
• "Plagiarism panic" and detector-based accusations.
• Handing students a tutor you haven't tested.

---

What's next: from the classroom to the finance team — AI for accounting and financial reporting.