Scalable Design Systems

Built from scratch for client products, then kept up to date inside Volvo's app with AI

I've built design systems alone from scratch, and kept components alive inside a large org including using AI to help maintain and ship the update.

Here's what the work looked like, and where I think system governance needs to go next.

Screenshot of a user interface for a vehicle rental or parking app, with menus, maps, and options for booking, payments, and vehicle management.

Screenshot of a website with a car on a road, showing a headline about used car purchases in Swedish, and navigation options, icons, and text describing how the service works.

Built alone

I owned the design system end to end for Golden Ridge (an early stage startup) and Trustie (an early incubation product by Länsförsäkringar) including tokens, type scale, icon library, component states.

Both products shipped on top of it.

A visual presentation of user interface design examples including score indicators, usage examples on a mobile device, color themes for light and dark modes, and implementation notes for a component.

Components shipped inside a larger system

At Volvo, I shipped and mainained two components without adding duplication in the system: a safety score view and a list view with score in Volvo’s shared component library.

I audited the existing system before building anything new. The component inherits same standard as the rest of the system: platform status, states, motion, accessibility, do's and don'ts.

A screenshot of a GitHub page titled 'chore: [DIS-1443] - Update SafetyCoach Score view'. The page includes a section with two circular score indicators showing '82', one in a lighter blue and the other in a darker blue, with an arrow between them. There is also a comment section with a user's comment, links for task tickets and UX specifications, and a list of recent update comments.

Keep it up-to-date with AI

When the design system pushed a new color and type standard, I used AI to implement the update to the score component's code, and reviewed the update with engineering team. The change was merged and shipped on iOS and Android.

What’s next

AI can already catch a lot of what review cycles exist to catch. It’s worth building a lightweight workflow using skills and evals so consistency gets checked and executed automatically, not in endless meetings and reviews.