Two directions in one course. Designing AI features people trust: confidence, grounding, human oversight, EU AI Act obligations. And using AI in your own work without letting plausible output pass for evidence.
Designers, product managers and engineers in the same room. AI features fail at the seams between these roles far more often than inside any one of them.
Legal and compliance colleagues are welcome for module 3, often the first time design and obligation are discussed together.
Show confidence honestly, and design what happens when the model is wrong.
Evidence, sources and traceability a sceptical expert can actually check.
Where approval belongs, and where a confirmation dialogue is just theatre.
Disclosure, oversight of high-risk systems and explanations in the interface, not the T&Cs.
Synthesis, prototyping and audit agents, with review norms that keep rigour.
A data boundary, agreed prompts and an escalation path for bad output.
Every module ends with work applied to the team's own product, not a case study from someone else's company.
Documented patterns for confidence, grounding, oversight and failure, specific to your product rather than generic examples.
EU AI Act and GDPR obligations expressed as design requirements a team can review against before release.
Where AI enters your own process, the data boundary it respects, and how output gets reviewed.
Not legal advice: the compliance module translates published obligations into design practice and should be reviewed alongside your own counsel.