Eir. A mental health companion that had to feel warm before it could be useful.

The problem

Mental health issues are on the rise, especially among young people, and standard services aren’t keeping up. The waitlists are long, and many who are struggling can’t put into words what’s wrong when they finally sit down with a professional. We wanted to design for the space between sessions. Not a substitute for treatment, but a companion: log how you’re feeling in seconds, see your own patterns over time, and provide the psychologist with the context you couldn’t articulate yourself, if you want to.

We named it Eir, after the Norse goddess of protection, mercy, and healing.

My role

I was the lead designer on a team of four. I was responsible for the design direction and for taking the app from wireframe to a finished prototype in Figma. We did concept development and testing together as a group.

The concept

The core loop is intentionally simple: choose one of five emojis for the day, and write a short post if you’d like. Everything else builds on that. A calendar that changes color based on your mood. An overview that shows how you’ve been feeling alongside data like activity and heart rate. An AI chatbot for the moments between human interactions, with a pathway to real healthcare professionals. And BankID at the door, because a medical record this personal must feel safe and secure.

The process, and what users caught us off guard with

We tested the wireframe with 8 people in October (10 were planned, but illness took out two): 12 tasks, 8 follow-up questions, 20 to 30 minutes each, mostly in people’s homes, because you get better feedback on your own couch than in a group room at school.

The most important finding was also the most painful. The flow for creating a new journal post, the very heart of the app, was the feature that performed the worst in testing. Five out of eight struggled with it. Our home screen had three different boxes that all appeared to start a post, and people didn’t realize they could change a selected emoji. If logging feels difficult, the app is worthless, so the entire flow was redesigned into a single, clear path.

Users changed more than that:

• Two of them pointed out that any login method other than BankID raised security concerns in such a sensitive app. They were right. BankID became the only way in.
• They asked for a profile page we had never planned. It’s there now, with streaks and weekly stats, because several people, independently of one another, wanted control over their own data.
• The “Stats page” got a new name. The testers thought the word was cold and impersonal for something so private, and suggested “Overview” instead. We agreed, and took it a step further: the overview became a scrollable path through your week, day by day, smiley face by smiley face, instead of a wall of graphs.
• The calendar became clickable. Select a day, read that day’s post.

There was one thing we deliberately didn’t change: the AI chat. Testers were neutral or skeptical about talking to an AI about mental health, but no one could point to a design problem. Our hypothesis was that a gray robot in a wireframe feels worse than a warm, clearly labeled assistant in a finished design. So we kept the feature, gave it a friendlier icon that’s upfront about it being a bot, added settings for how much feedback it provides and what it remembers, and postponed the decision until the next round of testing rather than redesigning based on doubt.

Warmth as a requirement

Halfway through, our first color palette failed the contrast checks. Fixing the accessibility issues forced us to completely rethink the palette, and the app ended up looking better for it: a calm blue paired with warm yellow and orange, the Montserrat font for readability, rounded shapes and soft shadows throughout, and a small sun made of gradients on the profile page. The contrast in the main text became clearer. The five emojis communicate through both color and expression, so they work for color-blind users as well.

WireframeFinal design

None of this is just for show. The whole point of the app is that someone having a bad day will open it anyway. Every soft corner serves a functional purpose.