AI project day at school: from an idea to a safe prototype

An AI project day is an intensive one-day format in which pupils use artificial intelligence to solve a specific problem — from initial design through a prototype to a presentation. The goal is not to learn how to operate a particular tool, but to experience the full cycle: identify a need, write prompts, verify outputs and decide what to do next.

Who a project day is suitable for

The format works at lower and upper secondary level. It fits into project weeks, computing and media literacy lessons, or a cross-curricular day. It does not require advanced technical knowledge — but it does require a teacher who is already familiar with AI (see How to introduce AI at school).

How to choose a real school problem or need

The strongest project days address a real need in the school or community: clearer web pages for first-year pupils, a guide for newcomers, recycling instructions or an anti-cyberbullying campaign. Avoid briefs such as “create anything with AI” — without a framework, pupils simply experiment with generating images.

• Ask teachers and the pupil council what is not working well at school.
• Choose 2–3 project briefs and let classes vote.
• State the problem in one sentence: “How can we make it easier for new pupils to find their way around our building?”
• Add the target audience, constraints (time, language, technology) and success criteria.

What pupils can and cannot enter into AI

• OK: general descriptions of a problem, public information about the school, sample teaching texts.
• OK: fictional characters, invented data for statistics exercises.
• Not OK: full class lists, parents' email addresses, health information.
• Not OK: copying entire copyrighted textbook chapters.

How to teach prompting, verification and hallucinations

The quality of AI output stands or falls on the quality of the instructions. Teach pupils the four-part prompt: role (“You are an editor of the school website…”), context (“We are writing for Year 9 pupils arriving from another school…”), task (“Suggest five FAQ questions”) and format (“Answer in bullet points, each no more than 2 sentences”).

Explain the term hallucination: AI can invent a source, an author's name, a statistic or a legal provision. Pupils must verify every factual claim used in their project against a trustworthy source (an official website, an encyclopaedia, a specialist article). Establish a rule: “If it has no source, it does not go into the prototype.”

What can realistically be produced in one day

Six lessons are enough for a design, a simple prototype and user testing. Do not expect a finished product — expect evidence that the solution makes sense. Typical outputs include a clickable website wireframe, an illustrated written guide, a short video, a poster design or a concept for a school AI guide.

Suggested structure for a six-lesson day

1. Lesson 1 — Introduction and ethics: rules for working with AI, data protection, hallucinations.
2. Lesson 2 — Exploring needs: interviews, a questionnaire, data collection.
3. Lesson 3 — Defining the brief and prompting: the four-part prompt, initial concepts.
4. Lesson 4 — Prototype: text, a wireframe, images, video or audio.
5. Lesson 5 — User testing: 3 participants from outside the team, gathering feedback.
6. Lesson 6 — Presentation and reflection: what worked, next steps, recommendations for the school.

Reflection for teachers

After the day, set aside 30 minutes in the staffroom. Ask: Where did pupils get stuck? Where did AI simplify the work, and where did it lead to superficial results? What will the school actually take forward from the prototypes? A brief written reflection (one A4 page) provides a foundation for planning future project days.