Course programme

Check an AI answer before you use it

A short demonstration for healthcare staff. Using fictional material, you see AI draft a message and learn to spot invented facts. You complete a quiz, try your own prompt and compare the answer with the source.

What you take away

A simple way to check AI answers and a worksheet. You also see what S1 covers. S0 does not lead to a certificate and is not required before S1.

Introduction to AI

Goal

Learn to use AI in clinical and research work: write precise prompts in Claude, ChatGPT and Gemini, check answers and sources, anonymise records and work with literature. We also cover patient re-identification risk, the AI Act and what the SynthID watermark does not tell you. The course is for physicians and researchers, including those new to AI.

What you practise

1. Prompt and source: specify the task, the material, rules for missing information and the output format, then compare answers from Claude, ChatGPT and Gemini.
2. Checking answers: find invented information, omissions, changed meaning and fake sources, then check the correction.
3. Anonymisation: in a synthetic visit note, identify direct and combined identifiers, remove unnecessary details while keeping the event and its consequences, and assess the remaining risk. Distinguish anonymisation from pseudonymisation.
4. Working with literature: draft an outline and a text from publications with AI, checking that the citations exist and say what the AI claims.
5. The AI Act: analyse a use case in a healthcare setting: purpose, users, oversight and AI literacy activities. Distinguish a training record from an assessment of organisational compliance.
6. SynthID: interpret example watermark-detection results. No watermark does not prove that a human wrote the text, and a watermark does not make it true. No access to a detector is required.
7. Independent work: complete a task on new material, check the output and receive feedback. Finish with a knowledge test.

Evening 1: source, prompt and data

• 01Platform warm-up and the essentials of language models.
• 02Source-based prompts, missing information and output checks in Claude, ChatGPT and Gemini.
• 03Anonymisation lab using a synthetic visit note.
• 04Preserving who said what and how certain they were, using a fictional dialogue.

Evening 2: assessment, literature and independent work

• 05Checking meaning, errors and sources. Working with literature: an outline and a draft.
• 06The AI Act and SynthID: two different questions about AI use and content provenance.
• 07Final task on new material and discussion.
• 08Knowledge test and your next step in practice.

Materials and certificate

You receive a prompt-building card, a data checklist, worked examples and your own exercise submissions. The materials stay available to participants who retake the test. Passing the test earns a completion certificate listing the programme. The trainer reviews the practical project separately, and we confirm the feedback deadline before the course. The test score alone does not mean the project has been reviewed.

The certificate can document the development of AI literacy. It does not certify an institution's compliance with the AI Act, and Article 4 does not require any specific certificate. The course does not carry professional education credits. More: European Commission guidance.

Organisation

All you need is a computer, a browser and internet access. We use our training platform and fictional material, with no real patient data. The anonymisation exercise does not approve real records for use in any particular tool. You do not need to buy an AI subscription beforehand. We confirm the terms of participation by email before taking payment.

Source for the content-provenance module: Google, SynthID Text.

Extension: AI in practice

Goal

Moving from “I can write a prompt” to “I have a working system for using AI”: advanced techniques, your own knowledge base, a documentation workflow, statistics from papers and from your own data, running an agent in practice, a mini AI policy and a team project for rolling AI out in an institution. You leave with a workspace set up in class, not a to-do list for later.

Learning outcomes

Knowledge

1. You know advanced prompting techniques (persona, few-shot, step-by-step reasoning, structured formats, chains) and when to use each.
2. You understand how an AI agent differs from a chat and where the risks of delegating to it lie.
3. You know what an institutional AI policy must contain and what Article 4 of the AI Act requires.

Skills

4. You build your own knowledge base and measure how it changes output quality.
5. You design a workflow: a repeatable process broken into a chain of prompts with verification points.
6. You read the statistics in papers with AI (p-values, confidence intervals, NNT, RR/OR) and verify the numbers. You run a simple analysis of your own data (from a CSV file to tables, a chart and conclusions) and check the results.
7. You produce finished documents with AI: a presentation, a spreadsheet and a letter, plus basic visual materials for the practice (business card, email signature, letterhead, patient leaflet), keeping to the anonymisation rules.
8. You run an AI agent on a task of your own, with human checkpoints.
9. You co-design a rollout plan for an institution: tools, rules, training and compliance documentation.

Programme

• M0Opening and technical warm-up
• M1Advanced prompting techniques
• M2Your own “My practice” knowledge base
• M3Documentation workflow
• M4Statistics and data with AI
• M5AI produces documents: presentations, spreadsheets, letters and practice materials
• M6Agents in practice
• M7Team work: rolling AI out in an institution
• M8A mini AI policy for my practice
• M9Wrap-up and rollout plan
• M10Open workshop (two-day format only)

What you take away

1. Your own “My practice” knowledge base (style, templates, glossary) set up in class.
2. A documentation workflow written out with verification points.
3. A set of finished documents: a presentation outline, a summary spreadsheet, a letter or template and a visual practice material.
4. A mini AI policy for your practice and a team rollout plan for an institution.
5. A certificate listing the programme and competences.

Vibecoding: your own tools without programming

Goal

In the workshop you build a working prototype of your own tool by talking to AI. You also learn where the line runs: what you may build and use yourself, and what needs a lawyer, a developer or medical-device certification. And why patient data never goes into prototypes.

Learning outcomes

Knowledge

1. You know what can realistically be built by talking to AI, what cannot, and which tools to use.
2. You know the legal limits: when a tool becomes a medical device (MDR), why a prototype must not process patient data, and when to bring in a specialist.

Skills

3. You set up a development environment from scratch, with no installation and no technical background.
4. You go through the full cycle (idea, description, prototype, new versions, bug fixes) entirely by talking to AI.
5. You describe a problem or bug to AI in a way that gets you a working fix.
6. You publish the prototype for your own use or as a demo.

Social competences

7. You judge your own tool critically: working does not mean fit for clinical use.

Programme

• M0Environment setup and technical warm-up
• M1What can be built: a map of options and examples to avoid
• M2Anatomy of the process: a shared prototype, live
• M3Workshop: your own prototype, first version
• M4Fixing bugs by talking to AI
• M5Workshop: second version and publishing
• M6Limits: law, data, responsibility

What you take away

1. Your own working prototype with a demo link.
2. A cheat sheet on reporting bugs to AI so you get a fix.
3. An “Is my tool a medical device?” checklist: warning signs and when to ask a lawyer.
4. A list of starter projects and description templates.
5. A certificate listing the programme and competences.

Publication workshop: Claude for Science

Goal

You leave with a working research environment: a project on your own topic, connections to publication databases and a manuscript outline under version control. You can run the whole cycle with AI: literature review with citation checks, statistical analysis, paper structure, writing, typesetting and versioning, verifying sources and disclosing AI use properly.

You work on your own topic throughout the course, not on examples, which is why the group is small.

Learning outcomes

Knowledge

1. You know which stages of research AI speeds up and where it is unreliable: choosing methods, interpretation, novelty.
2. You know journal policies and ICMJE guidance on AI: what to declare, why AI cannot be an author and who is responsible for the content.

Skills

3. You set up a research environment: a project for your own topic, connections to publication databases and organised materials.
4. You run a literature review with AI, checking every citation (does it exist, does it have a DOI, does it say what the AI claims), and build a literature matrix.
5. You run statistical analysis with AI and verify the results: matching the test to the data, interpreting the measures and checking the numbers AI reports.
6. You move from the matrix to the paper structure and write sections with AI in your own voice, without stock AI phrasing.
7. You typeset the manuscript in LaTeX / Overleaf (journal template, bibliography, tables and figures) with AI support.
8. You version your work so nothing is lost and versions for a supervisor or co-authors can be reproduced.

Social competences

9. You keep intellectual authorship: AI helps with the craft but is not a co-author of the thinking.

Programme

• M0Research environment setup
• M1Literature review with hallucination control
• M2From literature matrix to paper structure
• M3Statistical analysis with AI and result verification
• M4Writing with AI: introduction and discussion
• M5Typesetting the manuscript in LaTeX / Overleaf
• M6Versioning: an environment that keeps up with changes
• M7Ethics and disclosure of AI use in publications

What you take away

1. A configured research environment that stays with you after the course.
2. A verified literature matrix for your own topic.
3. A publication plan and a manuscript outline with bibliography.
4. A citation-checking checklist.
5. A template AI-disclosure section and a cheat sheet of publisher policies.
6. A certificate listing the programme and competences.