Three Mondays, 17 short videos, everything on this page. Watch before each session, bring your own project, and leave with a way of working you keep improving.
The first time, AI does not have to be faster. The gain comes from not throwing good thinking away: you capture it in a Skill, then make it a little better every time you use it.
About 4 hours across three weeks.
what makes project-based learning genuinely good, and where your own pedagogical model adds to the evidence.
an AI with a clear goal, task, constraints, output and quality criteria — and feed it your context first.
AI output against criteria and sources, compare alternatives, then use, revise or reject it.
a workflow that worked as a reusable Skill: instructions, context, quality checks and stop rules.
the right model and design system for the task, and automate the teacher work that repeats.
a controlled agent for formative feedback, with the judgement kept where it belongs — with you.
Brief it well, feed it your context, judge what comes back.
Capture a way of working that worked, so next time takes minutes.
Delegate a bounded, repeatable task — with a human checkpoint built in.
Throughout. Evaluating, iterating and human judgement are not a final module. They run through every practical video, and responsible, fair and ethical use is the floor under all of it.
Videos before a session, the session itself, a practical after it. Each block builds on the last: from one good brief, to a Skill you reuse, to an agent you control.
Why AI strengthens the teacher, what good PBL is, and where this training is heading.
5 videos
The learning outcomes, the route through the three sessions, and the core idea: AI does not replace pedagogical judgement. Responsible, fair and ethical use is the floor, not the ceiling.
Recognise the scientific core of strong project-based learning and complement it with the pedagogical model of your own school and practice. The principle the whole training rests on: poor pedagogical input gives poor AI output.
The three levels of working with AI — chatting, reusable ways of working, delegated work — and what each one asks of you. You will know where the training is building towards before you start.
The minimal AI toolkit you need for the training, and the ethical principles you apply from the first prompt on: privacy, bias, transparency, human responsibility and the rest.
Design from learning goals, student agency, scaffolding and evidence of individual learning — then decide, deliberately, what AI may take over and what it must not.
One good AI workflow, run live on a real teacher’s material.
Brief it, feed it context, judge what comes back.
3 videos
From a vague question to a clear prompt: goal, task, constraints, output and quality criteria. Prompt Cowboy as a tool to get there faster.
A reliable workflow, run end to end: desired output → find relevant knowledge → sources → a golden example → your own context → the prompt. The rubric case is the worked example.
Judge and iterate systematically: use your criteria, check the sources, compare alternatives, bring in a second model as a reviewer if it helps — then use, revise or reject.
The central whole skill: capture a workflow that worked as a Skill you will use again.
Skills, the right model, a design system, and the work that repeats.
4 videos
Capture a successful way of working with instructions, context, quality checks and stop rules, then improve the Skill on feedback. Also: external Skill libraries, and the risks of using one you did not write.
Compare models on quality, speed, context window, multimodality, price and the type of task — without treating any one model as the default for everything.
Make AI output consistent in style, structure and format — worksheets, slides, lesson material. Add your own design principles and UDL, and make the relevant Drive sources available to the AI.
Routines and scheduled tasks for the work that comes back every week: collecting current sources on a project theme, preparing periodic project updates, and more.
The culminating whole skill: a controlled AI agent for formative feedback.
Build the feedback agent at your own pace, then keep building.
5 videos
When is an agent actually useful, which steps does it get, and where must the human checkpoints sit?
Design the input, the sources, the assessment criteria, the Skill and the process.
Automate it, test it, find the failure cases, and improve it iteratively. The failure cases are the point.
A short gallery of PBL agent ideas worth building, across Design, Launch, Coach, Assess and Improve — with the method library as the place to start.
From loose experiments to a personal library of Skills, routines and agents that get a little better every time you use them.
0 of 16 files available so far. The rest follow as the videos are recorded.
| File | Type | Belongs to | Link |
|---|---|---|---|
| Training overview and route | Slides | Welcome: AI strengthens the teacher | follows |
| The PBL framework, one page | Reading | What makes PBL genuinely good? | follows |
| Responsible AI use: the principles card | Checklist | Tools and responsible AI use | follows |
| What AI may and may not take over: a decision sheet | Worksheet | Protect the learning | follows |
| Session 1 slides | Slides | Live session 1 | follows |
| Prompt structure template | Template | Write a good AI brief | follows |
| Context-first workflow checklist | Checklist | Context first: feed AI what is good | follows |
| Rubric case: the worked example | Example | Context first: feed AI what is good | follows |
| Use / revise / reject: the review sheet | Worksheet | Improve and test AI output | follows |
| PBL audit worksheet | Worksheet | Audit one of your own PBL projects | follows |
| Session 2 slides | Slides | Live session 2 | follows |
| Skill template | Template | Build your first AI Skill | follows |
| Model comparison sheet | Worksheet | Choose the right AI model | follows |
| Design system starter for lesson material | Template | Steer form and content with a design system | follows |
| Session 3 slides | Slides | Live session 3 | follows |
| Agent design canvas | Template | Build a feedback agent — part 1 | follows |
Forty problem-first methods across Design, Launch, Coach, Assess and Improve — the reference the training draws on, and the place to start when you want an agent idea after session 3.
Stuck, or something on this page is wrong? Tell your trainer in the session, or use the feedback form on any method page. This page is one perspective on the training, not a contract; the live sessions may adapt to what you bring.
Co-funded by the European Union under Erasmus+ KA210-SCH. No student data. No teacher material stored. Watched ticks stay in your browser only.