AI Boss Tools • Back to the Hub • adriandunkley.net/aiforschools
AI for Schools and Teachers
Your students already use AI. The serious question left is whether the adults in the
building understand it better than they do. I have taught AI weekly, free, since 2018,
trained thousands of Caribbean professionals, and written a framework any school can
adopt this term. This page collects everything a school needs in one place: the
framework, the classroom guidelines, a working prompt library, the research worth your
time, and my straight recommendations on what to adopt, what to pilot, and what to
refuse.
Everything else on this page hangs off one document: the AI Framework for Schools and
Teachers, an eight-module course I built from the strongest published guidance in the
field (AILit, TeachAI, the Australian Framework, UNESCO, NCEE) and from years of
teaching AI to people who had never touched it. It is free, self-paced, and written for
educators, not engineers.
Free Course • 8 Modules
The AI Framework for Schools and Teachers
Foundations, AI literacy, classroom practice, academic integrity, data privacy, equity, phased rollout, and governance. Each module ends with something you can use the same week: a policy draft, a lesson move, a rollout plan.
The article version: five pillars, a staged rollout you can start this term, the classroom moves that work, and the guardrails for integrity, privacy, and equity. Read this first if you want the argument before the course.
Every framework and study behind the course, collected with notes on what each is good for. Check my reasoning against the originals, or use the same sources to build a version that fits your school.
Distilled from the framework and from watching schools get this right and wrong. Adopt
them as written or argue with them, but decide something: a school with no position on
AI has taken a position, and it is the worst one available.
Teachers first, students second.
No student-facing AI until the teaching staff has used the tools long enough to predict where they fail. An educator who has watched a model invent a citation teaches verification with authority no policy document carries.
Disclosure is the default.
Students state what AI they used and how, on every piece of work, without penalty for honesty. The moment disclosure is punished, it stops, and you lose the only visibility you had.
Redesign the assessment, not the ban.
Any task a chatbot completes in one prompt was measuring recall, not thinking. Move the marks to process, oral defence, in-class production, and revision history, and the cheating question shrinks on its own.
Never feed student data to a public tool.
Names, grades, health notes, family details, photographs: none of it goes into a consumer chatbot, ever. Anonymise before prompting or do not prompt.
Respect the age floors.
Most consumer AI tools set 13 as a minimum, some 18. Those lines exist for legal and safety reasons, and a school that ignores them takes on liability it never needed to hold.
AI output is a draft, never an authority.
Teach every student the same reflex: generate, then verify against a source the model did not write. The skill of the decade is not prompting, it is catching the confident error.
Equity is a budget line, not a value statement.
If AI access depends on who owns a laptop or a subscription, the school is amplifying the gap it exists to close. Provide access at school or delay the rollout until you can.
Name an owner and a review date.
One named person owns the AI policy, and the policy carries a review date no more than a term away. The technology moves faster than any document; the review cadence is the policy.
The rollout that works starts with what students already do together, then adds the guardrails.
Working Tools
The Teacher Prompt Library
The eight prompts I use most in teacher training, built to save hours on the work that
eats evenings while keeping every judgement call with you. Copy one, paste it into the
AI tool your school has approved, and replace the bracketed parts. Verify everything
before it reaches a student; the last guideline above applies to teachers too.
Lesson DifferentiatorOne lesson, three levels of readiness, same objective.
You are an experienced [subject] teacher. Here is my lesson objective and outline: [paste]. Rewrite the core activity three ways: one for students working below grade level, one at grade level, one ready for extension. Keep the same learning objective in all three. For each version, state what I should watch for as evidence of understanding.
Rubric BuilderA rubric that rewards process, not just product.
Build a rubric for this assignment: [paste assignment]. Four criteria, four levels each. At least one criterion must assess the student's process (drafts, revisions, cited sources, or oral explanation) rather than the final product, so the rubric still works if a student used AI. Write level descriptors a 14-year-old can understand.
Misconception FinderPlan for the wrong answers before they arrive.
I am teaching [topic] to [age] students. List the six most common misconceptions students hold about this topic, why each one feels logical to a student, and one question I can ask in class that exposes each misconception without embarrassing whoever holds it.
Socratic Sparring PartnerFor students, supervised: a tutor that asks, never answers.
You are a tutor for a [age] student studying [topic]. Never give the answer or write any part of their work. Respond only with questions that move their thinking one step forward, and keep each question under 25 words. If they ask you to just tell them, reply with a smaller, easier question instead. Start by asking what they already know about [topic].
Feedback First PassFaster written feedback that stays in your voice.
Here is my marking standard: [paste rubric or notes]. Here is a student's work with all names removed: [paste]. Draft feedback in three parts: two specific strengths tied to lines in the work, the single highest-priority improvement, and one question that pushes the student's thinking further. Under 150 words, encouraging but honest. I will edit before sending.
Reading-Level ShifterThe same text, accessible to every reader in the room.
Rewrite this text for a reading age of [X] while keeping every key fact and the original meaning: [paste text]. Then list the vocabulary you simplified as pairs (original word, simpler word) so I can pre-teach the original terms to students who are ready for them.
Retrieval Quiz GeneratorLow-stakes quizzing with distractors that diagnose.
From this material: [paste], write a 10-question retrieval quiz for [age] students. Mix formats: multiple choice, short answer, and two questions that apply the idea to a new situation. For each multiple-choice item, make every wrong option a real misconception and tell me which misconception each option represents.
Parent Letter DrafterExplain your AI policy without the jargon.
Draft a one-page letter to parents explaining how our school uses AI. Cover: what students will and will not do with AI, how we protect their child's data, what we ask parents to reinforce at home, and who to contact with questions. Warm, plain language, no technical terms, reading age 12. Here are our policy points: [paste].
Read the Evidence
The Research Worth Your Time
The field produces more opinion than evidence. These are the sources I built the
framework on, with a note on what each one is good for. Read them yourself;
never outsource your judgement, not even to me.
The most rigorous definition of AI literacy in print: competences across engaging, creating, managing, and designing with AI. Use it to define what your students should know by each stage.
Policy language, sample guidance, and implementation checklists co-developed with education ministries and Code.org. The fastest route from no policy to a defensible draft.
A national framework built on six principles including teaching and learning, human and social wellbeing, transparency, fairness, and accountability. The best model for whole-system policy.
The global reference point, including its recommendation of age limits and human-centred deployment. Essential for the governance module of any school policy.
Independent risk assessments of the AI products students actually use, including age ratings and privacy reviews. Check any tool here before it enters a classroom.
Practical prompt libraries and teacher-facing training material. Good for staff professional development sessions once your policy exists.
Staff go first. A teacher who has watched a model fail teaches verification with authority.
My Position
Adopt, Pilot, Refuse
Schools keep asking me for a neutral summary of options. I decline. After training
thousands of people and watching AI rollouts succeed and fail across the region, these
are my recommendations, stated plainly so you can hold me to them.
Adopt Now
Teacher-facing AI for preparation. Lesson differentiation, rubrics, quiz generation, and parent communication, with the teacher editing everything. Lowest risk, highest immediate return.
Staff AI literacy training before any student rollout. Use the framework course or bring in training; either way, staff go first.
A written policy with an owner and a review date. One page beats none. Module 8 of the framework gives you the draft.
Disclosure-based integrity rules. Students declare AI use without penalty, and assessments are redesigned to reward process.
Pilot Carefully
Supervised student use in one subject, one term. Pick a willing teacher, define what success looks like before you start, and review honestly at the end.
Tutoring modes with guardrails. Socratic configurations that question rather than answer, used in class where a teacher can see the screen.
AI in assessment design. Run your existing assessments through a chatbot to find which ones it can complete; those are the ones to redesign first.
Refuse
AI detectors as evidence for discipline. They produce false positives, and they fall hardest on students writing in a second language. Never end a student's case on a detector score.
Student personal data in consumer tools. No name, grade, or family detail enters a public chatbot. This one is not negotiable.
Unsupervised AI companions for children. Conversational AI marketed as friendship has no place in a school. The wellbeing risk is documented and growing.
Any rollout that depends on home devices. If participation requires a family subscription, you have built inequity into the timetable.
Bring It to Your School
Policy on Paper Is Not Practice in Classrooms
I train school leadership teams, staff rooms, and education ministries across the
Caribbean, from a single staff development day to a full rollout with governance.
If your school is ready to move from reading about AI to teaching with it, write to me.
