I have taught AI in Jamaica for seven years, mostly outside classrooms: community centres, conference rooms and free Saturday sessions online, open to anyone. Thousands of Jamaicans have come through them. The question parents and teachers ask me most is when schools will teach this.
Not soon enough. Jamaica's formal system has no national plan for AI education, while Singapore, Estonia and China have been building theirs for years. Plenty of individual teachers are doing good work with little support. The problem is a system that has not yet decided AI literacy is its job.
This piece sets out what AI education could look like from primary school to university, using approaches that already work elsewhere and tools that cost nothing.
What Singapore and Estonia did that Jamaica has not
Both countries made a national decision first, then funded teacher training and gave schools a framework to adapt. Singapore layers the work: younger students learn what AI is and where they meet it, older students learn how it works, and the strongest students build systems. Basic AI literacy is expected of all students.
Estonia's ProgeTiger programme, running since 2012, brought programming and computational thinking to children from the start of primary school. Much of it is "unplugged": card games and group activities that teach how a machine learns from examples without a computer in the room. That matters for Jamaican schools where one computer lab serves the whole school.
A student who leaves sixth form in 2030 without basic AI literacy will enter banks, hotels, clinics, farms and ministries that already use these tools. The gap is not only in tech jobs.
Primary school, ages 6 to 12: learning by doing, mostly without screens
Primary AI education needs activities, not equipment. Children sort objects into groups and learn that this is what a classification model does. They guess the next item in a pattern and see how prediction works. They work out why a library system might suggest one book after another.
From about age 10, a browser is enough. Google's Teachable Machine lets a class train an image model with a webcam in about 15 minutes: show it examples, label them, and watch it learn to tell them apart. Machine Learning for Kids adds the same idea to Scratch, which many Jamaican primary schools already use.
Start with the experience and give the definition afterwards. A child who has watched a model confuse a mango with a ball understands "training data" better than one who copied the term from the board.
Secondary school, ages 12 to 18: three tiers mapped to CSEC and CAPE
Secondary schools can teach AI in three tiers without waiting for a new subject.
| Tier | Who | What they learn | Where it fits |
|---|---|---|---|
| 1. AI literacy | Every student | What AI is, where it shows up, what it gets wrong, who is accountable | Maths (probability), social studies (jobs), English (judging AI text), science (testing a claim) |
| 2. Practical use | CSEC IT students | Prompting, checking outputs, basic data handling, using AI for research and analysis | CSEC IT practical work and SBAs, within CXC's rules |
| 3. Technical foundations | CAPE Computer Science | How models are trained and evaluated, simple Python models, trade-offs in system design | CAPE units on algorithms and programming |
Tier one is the one most schools skip, and it is the cheapest. An English teacher asking a class to find three errors in a chatbot's essay on Marcus Garvey is teaching AI literacy.
University and beyond: AI in every faculty, plus shorter routes
UWI's computing programmes include AI content, and the IMPACT AI Lab I co-founded at UWI gives students research experience. That is not enough if AI stays inside the computing department.
- Business students need AI for forecasting and customer analysis.
- Medical students need to know how diagnostic tools fail, not only how they succeed.
- Education students need to learn to teach with and about AI.
- Agriculture and law students need precision farming and AI liability respectively.
Jamaica also needs routes that take months rather than four years: certificates and bootcamps built with employers, so the skills match real vacancies.
Teacher training is the constraint
None of this works without teachers who understand it, and most Jamaican teachers have had no formal AI training. They do not need to become engineers. A maths teacher needs enough probability to explain how a model predicts. An English teacher needs to recognise AI-written text, teach students to use writing tools honestly, and design assessments that still measure the student.
The free material is good. Elements of AI from the University of Helsinki gives a solid grounding in roughly 30 hours. MIT RAISE's Day of AI publishes complete lesson plans. Google and Microsoft both run free AI courses for educators. Teachers who join StarApple AI's free weekly sessions pick this up quickly. What they lack is protected time for professional development and a principal who counts it.
What CXC should change
CXC sets the CSEC and CAPE syllabuses for most of the English-speaking Caribbean, so its choices decide what schools teach. The CSEC Information Technology syllabus has no meaningful AI unit. CAPE Computer Science covers algorithms and data structures but does not treat machine learning as a topic of its own.
Syllabus revision is slow, and I understand the constraints. The minimum is an AI literacy unit in CSEC IT and a CAPE unit on machine learning, practical tool use and AI ethics. Neither needs a new syllabus. Schools do not have to wait for either: the current syllabuses leave room for AI projects and discussion, and some schools already use it.
Free resources, by use
| Need | Resource | Cost |
|---|---|---|
| Lesson plans | MIT RAISE Day of AI; AI4ALL; Google CS First | Free |
| Teacher learning | Elements of AI (about 30 hours); Google and Microsoft educator courses | Free |
| Hands-on tools, younger students | Teachable Machine; Machine Learning for Kids with Scratch | Free, browser only |
| Hands-on tools, CAPE students | Python in Google Colab | Free with a Google account |
| Background explanations | 3Blue1Brown on YouTube; Khan Academy maths and computing | Free |
| Local support | StarApple AI weekly sessions for teachers and students | Free |
Prompt templates for Jamaican teachers
Paste these into ChatGPT, Claude or Gemini and fill in the brackets. Check every output before it reaches a student.
I am a [subject] teacher in Jamaica teaching [grade level]. I want to introduce AI concepts
in my class without leaving the CSEC/CAPE syllabus. Suggest 4 lesson activities
that connect AI to [topic I am teaching now]. Each should take about [time] minutes
and need [no technology / one computer / students' phones].
Create a 45-minute lesson plan that introduces machine learning to [age group] students
in Jamaica using an unplugged group activity (no computers). Use examples from Jamaica
and the Caribbean. Include instructions, materials and how I will check understanding.
Draft an AI policy for a Jamaican secondary school. Cover: what students may use AI for,
what teachers may use it for, AI in assessments and SBAs, how students disclose AI use,
and what happens when the rules are broken. Assume students already use these tools.
I teach CAPE Computer Science. Outline a 4-week module on machine learning basics using
Google Colab. Include weekly objectives, exercises, a marking rubric and links to the
current CAPE syllabus objectives. Students know basic Python.
What to do next
- Principals: put AI on the agenda of the next staff meeting and adopt a one-page AI use policy before the next exam cycle. The third prompt above gives you a draft in minutes.
- Teachers of any subject: run one unplugged AI lesson this term. It costs nothing and takes 45 minutes; the Day of AI plans are a ready starting point.
- IT and computer science teachers: complete Elements of AI over a term (about 30 hours) and run Teachable Machine with one class.
- Ministry of Education: set a target of AI literacy content in the National Standards Curriculum and fund protected CPD hours for it, starting with one pilot school per region.
- CXC: publish a timeline for an AI literacy unit in CSEC IT and a machine learning unit in CAPE Computer Science, with clear rules on AI in SBAs.
- Parents: ask your child's school what its AI policy is. If there is none, that answer tells you where to push.
Frequently Asked Questions
Who decides what Jamaican schools teach about AI?
Two bodies. The Ministry of Education sets the National Standards Curriculum for grades 1 to 9, and CXC sets the CSEC and CAPE syllabuses examined in upper secondary. A school can add AI content inside either without permission, but only these bodies can make it a requirement, so that is where to direct a proposal.
Are there age limits on AI tools for students?
Yes. OpenAI's terms require ChatGPT users to be at least 13, with parental permission under 18, and most Google consumer accounts also start at 13. For younger pupils, use tools that need no student account, such as Teachable Machine run from the teacher's screen, or accounts managed by the school through Google Workspace for Education. Check each tool's current terms before a lesson, because they change.
Can a teacher reliably detect AI-written homework?
No. AI detectors produce false positives, and students who write in a second language or a formal style are flagged more often. Better checks are asking for drafts and notes, a two-minute oral follow-up on the work, and in-class writing for part of the grade. Set the rule on disclosure in advance so the conversation is about the policy, not a suspicion.
What does it cost a school to start teaching AI?
Close to nothing for the first year. Unplugged lessons need paper and cards, Teachable Machine and Colab need a browser, and the main teacher courses are free. The real cost is teacher time, roughly 30 hours for Elements of AI plus lesson preparation, which principals should count as professional development.
How do I check whether an AI lesson resource is any good?
Test it yourself before using it with a class. A good resource states the learning outcome, works with the equipment you actually have, and shows AI getting something wrong as well as right. Be wary of vendor material that teaches one product rather than the concept, and of anything that asks students to create accounts or share personal data.