1. First, we conducted semi-structured interviews with a diverse group of professionals, from general practitioners and nurses to specialists and innovation leads. These conversations revealed that while many professionals were curious about AI, they lacked time and found most existing training options too abstract or irrelevant.
2. Next, we designed and distributed a comprehensive survey for healthcare workers across Flanders. The survey helped quantify knowledge gaps, training preferences, and AI perceptions across occupations, age groups, and experience levels. We discovered that doctors and managers tended to report higher AI literacy, while nurses and paramedics had less exposure but a strong interest in tools that could ease their workload. Interestingly, younger professionals were not necessarily more AI literate than older ones, indicating that AI is still largely absent from formal healthcare curricula.
3. Finally, we held focus groups with stakeholders from hospitals, universities, and professional associations. These sessions gave us space to interpret the survey results together with those working in the field and add depth to the findings. For example, nurses told us they usually preferred to follow training during (paid) working hours, whereas doctors were more likely to attend evening sessions. Many participants also raised concerns about the lack of accredited courses and trustworthy, hands-on training materials. Ethics was rarely a top priority in terms of training needs, but came up strongly when discussing trust, bias, and legal uncertainty around AI in clinical use.

With a clear understanding of the audience, the next step is to define what learners should walk away with. These goals should be realistic, motivating, and anchored in the learners' daily experience.

VAIA makes use of a standardized set of learning goals, customized for learning about AI.

1. Work with a (specific) AI tool
2. Be inspired
3. Create a framework for AI (management, ethics, law, communication...)
4. Technical implementation of AI (programming)
5. Specialisation (become an AI specialist in your domain!)

💡These goals are used for all the courses in our training database and they can be filtered accordingly.

A good learning trajectory respects people’s time, attention, and energy. It should feel like a solution, not another problem to solve. When asked what they wanted, healthcare professionals told us that they are (not) looking for:

• Introductory AI modules tailored for healthcare professionals (85% interest).
• Topics like planning/rostering (highly valued by nurses), medical error reduction, and decision support.
• Low interest in coding or technical development; high interest in facilitating AI use in one’s own field.

We also polled format preferences. Here we find quite notable differences between the professions:

• Online modules with flexibility for doctors.
• In-person sessions during working hours for nurses.
• All professions prefer co-created courses with universities and hospitals - not just tech vendors.

We are currently using these findings to co-develop training materials with input from healthcare professionals themselves. The EU AI Act, that came into effect in 2025, requires employers to ensure “sufficient AI literacy” among their staff. This adds momentum and urgency to get it right.

What will VAIA do now?

• contact hospitals, professional associations, learning providers, vendors and experts by experience to set up new training courses together
• Initiate accreditation processes in collaboration with professional associations
• continuous feedback loops from learners to refine the training
  

Collaborate and co-create

No one can develop a meaningful trajectory alone. Collaboration with users, educators, institutions, and policymakers is essential, not just for content validation, but also to gain trust and adoption.

Plan for change

A learning trajectory is never finished. As technologies evolve, so should learning trajectories. What will happen after the pilot phase? How will new content be added? Who maintains the platform? How will success be measured?