AI‑native UX for Misi
From an idea to the App Store: the first design of an AI colour-analysis app, built so people trust the result.
Client
Misi App (Consumer AI)
Role
First product designer
Year
2024
Services
End-to-end UX/UI · Onboarding · AI result design · Mobile design system

01
Context
Misi is a consumer app for personal colour analysis: it tells you which colours suit you, then turns that into outfits, make-up and shopping. When I joined, Misi was an idea. As the first designer, I worked with the founder and the developers to design the first version end to end — from sign-up to a personal palette — and it shipped to the App Store.
02
The algorithm worked. The trust didn't.
Colour analysis is expert knowledge. A colourist looks at your eyes, hair and skin in daylight and knows what to look for. Misi had to let people do this on their own, on a phone, with no training — and then trust a result that tells them which colours to wear.
For users: answer questions they've never been asked ("Is your undertone warm or cool?") without guessing.
For the business: turn a one-time result into something people come back to — palettes, outfits, make-up, shopping.
An AI result people can’t question is a result they won’t trust — so every recommendation had to explain itself.
03
Constraints
No colourist in the room. Everything an expert would see had to become a question a non-expert can answer.
Phone photos lie. Light, make-up and hair change how colours look on camera.
Shopping isn't everywhere. Some features weren't available in every country, and the app had to say so without feeling broken.
A small team. The founder, the developers and one designer — every screen had to be buildable from shared components.
04
Translate the expert's method into small, visual steps.
The self-assessment follows the order a colourist works in: eyes, hair, skin tone, undertone, then how you like to dress — confidence and contrast. Each step asks one thing, shows real examples instead of descriptions, and moves a progress bar forward so the effort feels finite.
Before the camera, the app prepares the photo: remove make-up, pull your hair back, and only then ask for camera or gallery access — with a sentence on why.
The result isn't a label. "You are light spring" opens into what it means: your characteristics, your palette, colours to avoid, and for every colour, combinations, make-up and where to shop it.

The whole app, mapped with the founder and developers: onboarding, home, My Type, make-up and the palette.
05
Key decisions
1. Show, don't describe
2. Ask about the person, not just the face
3. Prepare the photo before asking for it

Preparation first, then permission — with one sentence on why the camera is needed
4. A result you can explore

Neutrals, accents, basics — and My Colour Dimensions: hue, value and chroma on a simple scale, one sentence each

The same palette applied to make-up, product by product
5. Let the colour be the interface

Combinations, before → after: from small swatches with captions to full-width colour. Colours to avoid follows the same pattern.
The design system
06
The first version shipped to the App Store, where Misi is live today as "misi Color Analysis & Style AI" — it has since grown into a broader style platform. misi.app ↗
The same problem shows up in B2B: when software makes a recommendation, people need to see why before they act on it.
“Her work is professional, quick, and full of ideas… she is up there with the best UX designers I have worked with.”
— Matthew Johnson, Founder, Misi App
on the App Store


