A business web app redesigned sprint after sprint

A solid base, a heavy experience

The existing application was functionally rich, with quality structured data and about 3,500 active files. But its use remained administrative: menus, sub-menus, forms.

Its users are agency managers, home-care aides in the field, nurses and advanced practice nurses. Often on the move, under time pressure, with very varied profiles and rights. The ambition: make it a reference product that can be marketed in the health and social care sector.

A design sprint, repeated for each feature

• An idea comes up from the client or users.
• A workshop or a user interview scopes the need.
• A design iteration answers that need.
• A user test, if needed, validates the proposal.
• A trade-off with the development team decides feasibility.
• The feature is developed in a sprint.

Cycle repeated 49 times since the project started.

The whole application, designed for the field

• Redesign of the entire application, with more than 70 screens.
• New key business features: handovers between professionals, a personalised care plan (PPA), a first for the sector since no HAS reference framework existed yet for this kind of tool.
• Regulatory compliance under way: Ségur, Pro Santé Connect, the national health record, HAS, GDPR, AI Act.
• A contextual AI assistant built into the journeys, designed with the AI team and our technical AI partner: I handled the visual side and the UX/UI journeys. The client took the bet of relying on AI to build the care plan, which the assistant generates directly from a plain-language request, ready to review and complete.

An honest miss, on the action-plan form

Over a certain period, the client became reluctant to involve users in our workshops: rarely available, they often cancelled. The one time we did work on the action-plan form with them, the result didn't convince them: we ended up with a long, unintuitive version, trying to cover every case rather than the essentials — an action plan first needs to be relevant and quick to fill in to be usable by every profile.

For lack of time, we built and shipped it as it was. No one used it. The form rested on no existing HAS framework: we had no solid ground to stand on, which made the mistake easier to make. Another signal I should have read better: the users involved didn't dare contradict a version designed by the head of the company, which skewed their feedback — I should have flagged that far earlier. I took a concrete lesson from it: assert my choices more, scope the need better upfront, and not hesitate to raise the alarm when needed. A few months later, I proposed a redesign of the form that did convince.