How to Design AI Professional Development for Teachers

Last reviewed: October 2026.

Most AI training for teachers is a single session: a tour of tools, a few impressive demos, and a room of people who go back to their classrooms and change very little. The problem is not the teachers. It is the design. Professional development changes practice when it is sustained, built around real tasks, and leaves people with something they can use the next morning.

This guide sets out how I design AI professional development for teachers, trainers and education leaders, drawing on programmes I have designed and delivered: the Future Educator programme with KFAS in Kuwait, a 24-week internal AI programme at Cambridge, workshops for civil society leaders with the African Union’s ECOSOCC, the HP Cambridge EdTech Fellowship, and the Designing Assessments with AI course for the Cambridge Assessment Network.

Why one-off AI workshops rarely change practice

A single workshop can build enthusiasm, but enthusiasm fades when teachers return to full timetables. Professional development fails less from a lack of initial interest than from the difficulty of sustaining engagement over the months that real skill development takes. AI adds a second problem: the tools change constantly, so training built around one product’s interface ages quickly.

The programmes that work share a different starting point. They begin with pedagogy, not technology. In Kuwait, day one was not about tools at all. Participants designed their own frameworks for what gets offloaded to AI, what the gains are and what risks need managing. The questions that mattered were: what do I want my learners to learn, where does AI help with that, and where might it get in the way?

Start from a competency framework

A shared framework stops AI training becoming a collection of tips. UNESCO’s AI competency framework for teachers, published in 2024, is the most useful global reference.^[1] It sets out 15 competencies across five dimensions, a human-centred mindset, ethics of AI, AI foundations and applications, AI pedagogy, and AI for professional learning, organised into three progression levels: Acquire, Deepen and Create.

I use it in two ways. First, to check balance: a programme that is all foundations and applications, with nothing on ethics or pedagogy, will produce confident tool users rather than better teachers. Second, to set realistic goals: most teachers in a first programme are working at the Acquire level, and the design should reflect that rather than aiming for Create in a week.

Design principle 1: little and often

For the internal programme I designed at Cambridge, I used what I call the 30-minute formula: twice-weekly sessions of 30 minutes over 24 weeks, each split into ten minutes of instruction, ten of live demonstration and ten of hands-on practice. Thirty minutes fits into a lunch break, and an hour a week over six months is far more sustainable than multi-day workshops that disrupt operational work. The full design is in Designing a 24-week AI training programme, and the opening module is written up in AI fundamentals for educators.

The same instruction, demonstration and practice structure works at any length. I used it for the UNICEF ministry workshop in Helsinki and in most of my workshops since. The aim is that every participant leaves having built something, not only having watched a presentation.

Design principle 2: blend online and in person

The Kuwait programme combined three months of structured online learning with in-person workshops at Kuwait University. Fifty STEM teachers completed it. Because participants arrived with a shared foundation from the online modules, the face-to-face days could focus on application, discussion and collaboration rather than covering basics. The final in-person days moved through pedagogy, safe and effective use, assessment, and action planning with peer presentations. The design is set out in the Future Educator case study.

The HP Cambridge Partnership for Education EdTech Fellowship follows a similar logic for senior leaders: a five-month hybrid programme with an annual residency week in Cambridge, where I facilitate the practical AI and implementation thread. See the EdTech Fellowship case study.

Design principle 3: practise on real tasks

Teachers adopt AI when it saves planning time without removing their professional control. So the useful units of practice are the tasks teachers already own: lesson planning, assessment, differentiation and classroom activities. In the November 2025 Kuwait workshop, teachers scored their own tasks with a four-question test: is it time-intensive, repeatable, scalable and expertise-based? Each teacher left with one repetitive task scored, a complete prompt built with Kuwait curriculum context, and peer feedback.

Simple, adaptable prompts help teachers get started without special setup. In 10 simple prompts for Cambridge teachers I share prompts for syllabus guidance, lesson planning and activities that I tested across ChatGPT, Claude and Gemini, which are similar enough that teachers can use whichever they have access to. The habit that matters is checking and verifying the output.

The same principle applies beyond schools. At the two-day ECOSOCC and Cambridge workshop in Nairobi in July 2026, civil society leaders worked on fundraising, project management and reporting tasks from their own organisations, using AI tools and then judging the output. More than 1,200 organisations applied for 50 places, which says a lot about the demand for practical training. See the ECOSOCC case study.

Design principle 4: keep judgement and ethics in the activity

Ethics lands better as part of prompting, review and learner-facing tasks than as a separate lecture at the end. In every programme I design, the AI drafts and the professional decides. In Kuwait we used a simple Human-AI Partnership Framework, Define, Draft, Refine, Validate, Deliver, so that review was built into every workflow.

One activity I now use early in most programmes asks participants to work with AI in their own home language and request culturally specific content. It shows both how capable the tools are and where their blind spots lie, and it builds a more realistic view of AI from the first session.

Assessment needs particular care. The Designing Assessments with AI course I designed for the Cambridge Assessment Network is built on generate, then judge: three live online sessions, six CPD hours, in which practitioners generate assessment items from structured inputs, evaluate them, and leave with a checklist for AI-generated assessment content. I have delivered four cohorts. The wider approach is in Assessment in the AI era.

Design principle 5: build a local multiplier

A first cohort is not a system. In Kuwait, ten teachers were selected for a train-the-trainer phase so that the practice could spread beyond the first group. That path is the difference between a successful course and a real chance of wider school use. Any programme funded at system level should plan for it from the start: who will train the next cohort, with what materials, and with what support.

Measuring whether practice changed

Attendance and satisfaction scores tell you people turned up and enjoyed it. They do not tell you whether teaching changed. Better signals are the artefacts participants produce and keep using:

lesson plans, assessments or resources created in the programme and used afterwards;
action plans presented to peers and followed up after a set period;
changes in how teachers describe their own judgement about when to use AI and when not to;
school readiness measures over time, such as those in the AI readiness framework.

Confidence matters too, but as a professional outcome rather than a mood: people need a method they can defend to colleagues. For schools working out where to begin, the AI readiness framework includes staff capacity as one of its five dimensions.

A blueprint you can adapt

Before: agree goals against a competency framework, survey participants’ roles, tasks, languages and connectivity, and choose tools teachers can access free.
Foundations: short sessions on how AI works, what it does well and badly, and responsible use, with the home-language activity early.
Practice: real tasks from participants’ own work, using the instruction, demonstration, practice structure.
Application: participants build and test something for their own classroom or organisation, with peer critique.
Multiplier: select and support local trainers, with materials they can adapt.
After: follow up on action plans and artefacts, not just attendance.

Teacher capability also sits at the centre of the other guides on this site: AI literacy in African schools and Sovereign AI for education systems. I also discussed what teachers need in South Africa on the first Lefa AI panel; my notes are in What AI means for Mzansi’s classrooms. To discuss a programme for your teachers or leaders, see Speaking or get in touch.

Sources

Niall McNulty is Director of Product and Innovation, Sub-Saharan Africa, at Cambridge University Press & Assessment.