What you will learn
Signature framework
SATHI — a structure for every prompt
The course's core skill: turning a vague request into five deliberate slots, so the AI builds what your classroom actually needs.
Board, class, subject, chapter and what was taught last — the context the AI cannot guess.
Language level, class size and the mix of learners the material must serve — described by need, never by label.
The exact artefact, its format and its time budget — a 40-minute plan, a 10-item quiz, a bilingual explanation.
Privacy red lines, source preferences and your verification plan — no real student data, facts checked against the textbook.
Ask the AI to flag its uncertainty, then change one thing at a time — prompting is a craft, not a single shot.
SAFE — review every output before classroom use
What you'll build
In the capstone you build a documented AI toolkit for one real classroom problem of your own — six domain prompts, a SAFE review, a reflection, an implementation plan and the no-student-data declaration — scored on a seven-dimension rubric.
See the capstone120+ ready-to-use prompt templates
A reviewed, searchable prompt library filterable by board, class, subject and language — each template ships with variables, an example, failure modes and a SAFE checklist.
Course Syllabus
Prepare to use AI deliberately, safely and with clear human accountability: what generative AI can and cannot do, the privacy red lines, and the SATHI and SAFE frameworks used throughout this course.
Learning Outcomes
- Explain, with classroom examples, what generative AI can and cannot do (LO1).
- State the privacy red lines: information that must never be entered into an AI tool, and why the teacher remains accountable (LO15, LO17).
- Apply the SATHI prompt framework and the SAFE output review to a simple teaching task (LO3, LO18).
Lessons
What AI Can and Cannot Do in a Classroom
Objective: Explain what generative AI can and cannot do in a classroom, and why fluent output still requires teacher verification.
Responsible AI, Privacy and Teacher Accountability
Objective: State the privacy red lines for classroom AI use and explain why accountability for AI-drafted content stays with the teacher.
Introduction to SATHI and SAFE
Objective: Apply the SATHI framework to build a complete classroom prompt and run the SAFE review on its output.
Module Assessment
Scenario Classification · 8 Questions
Visual Concepts
Flowchart
Teacher-Controlled AI Workflow
Checklist
SATHI Prompt Framework Card
Resources
Responsible AI Quick-Start Guide for Teachers
One page: the red lines, the teacher-review duty, the SATHI and SAFE cards, and the 12-situation classification activity.
Turn curriculum outcomes into precise prompt specifications, generate complete lesson sequences and explanations, and improve drafts through structured teacher review — with worked examples from Class 5 EVS to Class 11 Biology.
Learning Outcomes
- Convert a curriculum outcome into a SATHI prompt specification with board, class, prior knowledge and time constraints (LO2).
- Generate a complete lesson sequence — hook, explanation, modelling, guided practice, closure — and adapt it to a real period length (LO5).
- Run a structured review pass that turns an AI draft lesson into a classroom-ready plan (LO11, LO13).
Lessons
Turning Curriculum Outcomes into Prompt Specifications
Objective: Convert a syllabus outcome into a complete SATHI prompt specification with time, materials and level constraints.
Generating Lesson Sequences and Explanations
Objective: Generate a time-boxed lesson sequence and a level-controlled explanation for a real upcoming period.
Improving a Draft Lesson through Teacher Review
Objective: Run the four-pass review on an AI lesson draft and produce the corrected version through iteration.
Module Assessment
Curriculum-aligned Prompt Task · 6 Questions
Visual Concepts
Comparison Chart
Prompt Anatomy Diagram
Timeline Visual
Lesson Sequence Arc
Resources
Lesson-prompt builder and lesson-quality checklist
Fill-in SATHI builder for lesson prompts, the four-pass review checklist, and all six worked examples (Class 5 EVS to Class 11 Biology) in full.
Design valid assessments with AI: the blueprint matrix, four objective item types, five rubric and feedback formats, and the six-defect audit that catches ambiguity, wrong keys and bias before students do.
Learning Outcomes
- Specify assessment prompts through a blueprint: outcome, cognitive level, item type, difficulty, marks and evidence (LO6).
- Generate the full assessment toolkit — items, rubrics, checklists, self- and peer-assessment — with observable descriptors (LO6).
- Audit AI-generated assessments for the six defect classes and correct them through iteration (LO11, LO12).
Lessons
Designing Valid Assessment Prompts
Objective: Design an assessment through the blueprint matrix and prompt for the four objective item types in board format.
Creating Questions, Rubrics and Feedback
Objective: Generate an observable-descriptor rubric with matching self- and peer-assessment tools, and anonymised feedback stems.
Reviewing Assessment Quality and Bias
Objective: Find all six defect classes in a flawed AI-generated test and produce the corrected paper through iteration.
Module Assessment
Assessment-design Artifact (rubric-scored) · 7 Questions
Visual Concepts
Comparison Chart
Assessment Blueprint Matrix
Cycle Diagram
Assessment-to-Feedback Loop
Resources
Question-paper template, rubric builder and assessment audit sheet
The blueprint matrix template, rubric builder grid with observability rule, the six-defect audit sheet, and the planted flawed test with its corrected version.
Plan for learner variability with AI: the Differentiation Ladder, UDL in prompt form, accessibility features you can request by name, the five-version adaptation canvas — and the audit that stops deficit-based or stereotyped adaptations.
Learning Outcomes
- Generate same-goal tiered versions of one lesson along the Differentiation Ladder (LO7).
- Request accessible formats by name: readable layout, dyslexia-friendly structure, alt text, accessible tables, language support (LO7).
- Detect and correct the five deficit/bias failure patterns in AI adaptations (LO11).
Lessons
Planning for Learner Variability
Objective: Write differentiation prompts that fix the goal and vary the Ladder rungs, using needs-based descriptions only.
Creating Accessible and Differentiated Materials
Objective: Generate an accessibility-first version using named features and complete the five-version adaptation canvas.
Preventing Deficit-Based or Biased Adaptations
Objective: Detect all five deficit/bias failure patterns in an AI adaptation and repair them through one named-finding iteration.
Module Assessment
Scenario Differentiation Plan · 6 Questions
Visual Concepts
Flowchart
Differentiation Ladder
Checklist
Five-Version Adaptation Canvas
Resources
Inclusive adaptation canvas and accessibility checklist
The five-version adaptation canvas, the named-features accessibility checklist, the five-pattern bias audit, and the worked Class 5 repair example.
Use AI to name the wrong model behind consistent errors, build confront-rebuild-practise remediation, and design rechecks that prove whether it worked — across fractions, equations, force, photosynthesis, grammar and history, without ever profiling learners.
Learning Outcomes
- Distinguish misconceptions from careless errors using anonymised response patterns, and name the underlying wrong model (LO8).
- Generate four-part targeted remediation — confrontation, rebuild, worked example, scaffolded practice — for a named misconception (LO8).
- Design rechecks with transfer items and old-model distractors, and pre-commit the teaching decision for each outcome (LO8, LO18).
Lessons
Recognising Patterns of Misunderstanding
Objective: Distinguish misconceptions from mistakes and write a diagnosis prompt that names candidate models and separating diagnostics.
Creating Targeted Explanations and Practice
Objective: Generate and verify a four-part remediation pack — confrontation, rebuild, worked example, failure-point practice — for a named misconception.
Evaluating Whether Remediation Worked
Objective: Design a recheck with new surface, old-model distractors and transfer, and pre-commit the teaching decision for each outcome.
Module Assessment
Remediation Pathway with Justification · 6 Questions
Visual Concepts
Cycle Diagram
Remediation Cycle
Comparison Chart
Six Misconception Families
Resources
Misconception analysis worksheet and remediation planner
The model-naming worksheet, four-part remediation planner, recheck design grid, decision table, and all six domain examples worked in full.
Generate inquiry stimuli, project frames, stories, debates and simulations with AI — with the AI Support Continuum marking exactly where assistance must stop so the thinking, creating and concluding remain the students' own.
Learning Outcomes
- Generate inquiry stimuli — anomalies, provocations, mysteries, question ladders — that open student thinking rather than answer it (LO9).
- Design format frames for projects, stories, debates, simulations and design challenges, with roles, constraints and success criteria (LO9).
- Protect originality and agency through task design: local anchors, visible process, explicit allowed-use levels (LO18).
Lessons
Generating Inquiry without Replacing Learner Thinking
Objective: Generate inquiry stimuli with the no-conclusions boundary so investigation and conclusions remain student work.
Designing Projects, Stories, Debates and Simulations
Objective: Generate a complete format frame — including the interdisciplinary project — with roles, constraints, criteria and the student-work boundary.
Protecting Originality and Student Agency
Objective: Redesign an AI-vulnerable task with local anchor, process marks, checkpoint and an explicit allowed-use level.
Module Assessment
Inquiry-activity Design Task · 6 Questions
Visual Concepts
Timeline Visual
AI Support Continuum
Checklist
Five Format Frames
Resources
Project prompt canvas, discussion prompt deck and creativity safeguards guide
The five-frame canvas, twelve discussion/debate starters, the four-lever safeguards guide with the allowed-use card, and the worked interdisciplinary project.
Reduce routine workload safely: the green/amber/red task sorter, slot-based templates for notices, agendas, parent communication and planning, and the pre-send review — with an absolute red list of what never enters an AI tool.
Learning Outcomes
- Classify administrative tasks green/amber/red and apply the anonymise→generate→re-personalise routine to amber work (LO10, LO18).
- Build reusable slot-based templates for recurring professional documents in a consistent personal tone (LO10).
- Run the five-point pre-send review and transform unsafe prompts into privacy-preserving workflows — including the no-AI verdict (LO15).
Lessons
Reducing Routine Work Safely
Objective: Classify administrative tasks green/amber/red and apply the anonymise-generate-repersonalise routine to amber tasks.
Drafting Professional Communications and Documentation
Objective: Create reusable slot-based templates for two recurring documents, with a standing tone paragraph.
Reviewing Administrative Outputs
Objective: Apply the five-point pre-send review and transform unsafe prompts into privacy-preserving workflows, recognising the no-AI verdict.
Module Assessment
Administrative Workflow with Privacy Analysis · 6 Questions
Visual Concepts
Checklist
Administrative Privacy Red Lines
Flowchart
Green/Amber/Red Task Sorter
Resources
Teacher productivity prompt pack
All ten document-family templates with slots, the green/amber/red sorter card, the five-point pre-send review, and the six-case unsafe→safe drill set.
Create bilingual materials that preserve meaning, terminology and reading level: parallel generation, the teacher-owned terminology policy, back-translation and register checks, the Multilingual Quality Wheel — and the repair drill that trains the reviewing eye.
Learning Outcomes
- Generate parallel bilingual materials with an explicit teacher-owned terminology policy and reusable terminology tables (LO14).
- Apply the three parity checks — back-translation, terminology verification, register read-aloud — to AI translations (LO11, LO14).
- Review bilingual materials across the full Quality Wheel, including cultural fit, regional variation and script accuracy (LO14).
Lessons
Creating Bilingual Instructional Materials
Objective: Generate parallel bilingual materials with a teacher-owned terminology policy and a reusable terminology table.
Preserving Meaning, Terminology and Reading Level
Objective: Apply back-translation, terminology verification and register checks; produce rule-identical bilingual assessment instructions.
Reviewing Cultural and Linguistic Accuracy
Objective: Review bilingual materials across all seven Quality Wheel spokes and repair the five planted error classes through iteration.
Module Assessment
Bilingual Artifact with Review Log · 6 Questions
Visual Concepts
Radar Chart
Multilingual Quality Wheel
Comparison Chart
Terminology Decision Table
Resources
Bilingual prompt builder and translation-review checklist
The parallel-generation builder, terminology table template, three-check review card, seven-spoke wheel, all six worked bilingual examples, and the five-error repair drill.
Diagnose weak prompts to their cause, verify outputs through the eight-step decision tree — claims, sources, age fit, bias, objective — and iterate deliberately with version trails: the discipline that turns prompting from luck into craft.
Learning Outcomes
- Diagnose weak output to its cause — ambiguity, missing context or overconstraint — and fix the responsible SATHI slot (LO13).
- Run the eight-step Output Verification Decision Tree, including independent citation checking and bias review (LO11, LO12).
- Iterate deliberately: one change per version, findings as input, version trails as evidence, and a principled stopping rule (LO13, LO18).
Lessons
Diagnosing Weak Prompts and Weak Outputs
Objective: Diagnose weak output to its disease and fix the responsible SATHI slot, using the ten-point scorecard.
Checking Accuracy, Bias and Source Quality
Objective: Run the eight-step verification tree on cited content, independently checking every claim and reference, and record the verdict.
Iterating Systematically
Objective: Run a deliberate three-version comparison with recorded trail, select with justification, and apply the stopping rule.
Module Assessment
Prompt-debugging Challenge · 7 Questions
Visual Concepts
Flowchart
Output Verification Decision Tree
Comparison Chart
Prompt Disease Diagnostic Map
Resources
SATHI Prompt Quality Scorecard and SAFE Output Review Sheet
The 10-point scorecard, the SAFE review sheet, the eight-step verification tree card, the planted-citation passage, and the three-version debugging challenge.
Everything, assembled: choose a real recurring classroom problem, design its teacher-controlled workflow, build and test six prompts across the course's domains with version-trail evidence, and submit the documented toolkit — SAFE review, reflection, implementation plan and the no-student-data declaration — against the seven-dimension rubric.
Learning Outcomes
- Select a qualifying classroom problem and design its teacher-controlled AI workflow with the stops-here line marked (LO16, LO18).
- Build and test six domain prompts serving one workflow, with version trails and known failure modes as evidence (LO16, LO13).
- Document the toolkit professionally: SAFE review, honest reflection, term implementation plan and the signed declaration (LO15, LO17).
Lessons
Select a Classroom Problem and Design the Workflow
Objective: Select a qualifying problem and design its nine-step workflow with the stops-here line and module mapping.
Build and Test the Toolkit
Objective: Build, test and iterate all six toolkit prompts with version trails and recorded failure modes.
Document Evidence, Safeguards and the Implementation Plan
Objective: Assemble, self-score and (if needed) revise the complete capstone submission: SAFE review, reflection, plan, declaration.
Module Assessment
Capstone Toolkit Submission (7-dimension rubric) · 5 Questions
Visual Concepts
Checklist
Capstone Toolkit Canvas
Flowchart
Teacher-Controlled Workflow Map
Resources
Capstone toolkit template and 7-dimension rubric
The six-section submission canvas, the scoring rubric with level descriptors, the declaration text, and Sunita's completed fictional exemplar.
AI can draft, but it does not understand or verify. You remain responsible for the accuracy, fairness, privacy and classroom-appropriateness of anything you use.