Ian Bone
Case study · 03 · 4 months

AppliedIntuition

Turning a standard design system into a multi-brand, agent-ready one, with semantic tokens, enforced rules and a grammar that teaches a model how to design with it.

Role
Senior Product Designer, via Orium
Timeline
4 months
Disciplines
Design systems · Semantic tokens · Multi-brand · Agentic design
An illustrated car interior on a city highway, the dashboard screen showing a driver-assist view
  • Design systems
  • Semantic tokens
  • Multi-brand
  • Agentic design
Overview

This project came out of design system work I did for NordicTrack. The stakeholder I worked with there was happy enough with it that when he moved to Applied Intuition, he reached out to see if I would come and support their team in building out an agentic design system. It has grown through a number of extensions and new scopes of work, but it has stayed one project, about four months long, focused on expanding the depth of coverage their design system has.

01

Before

A standard design system, without the rules to hold it together.

Applied Intuition started with a fairly standard design system that lacked the polish of mature systems like Atlassian or Polaris. Nothing stopped someone from building something that looked off-brand, so a person or an agent coming into it had no reliable way to make a product that felt consistent and looked like it came from Applied.

  1. Raw values everywhereRaw values and primitives were used directly all over the place, so nothing tied a colour or a size back to what it was for.
  2. No semantic tokensThere was no set semantic token system to name the job a value does.
  3. No rulesThere was nothing written down about how the system should and shouldn’t be used.
  4. No enforcementThere were no linters to catch someone using the system in a way that broke consistency.
02

System

Semantic tokens, multi-brand layers, and a layer agents can read.

I spent most of the project refining, building and testing new semantic token build-outs across colour, size, spacing and motion. On top of that we added multi-brand layers, so Applied could sell the design system white-labelled to brands, mostly in the car industry, for their software systems.

  1. Agent layerDocumentation and grammarWritten so agents can read the system and properly understand how to use it.
  2. Brand layersWhite-label themesOne system that can be re-skinned for each brand that licenses it.
  3. Semantic tokensColour, size, spacing and motionEvery value named for its job, so components stop reaching for raw values.
  4. FoundationThe existing primitivesThe raw values the system started with, now one layer down.
03

Grammar

Codifying what an experienced designer knows.

The biggest challenge was figuring out how to take my knowledge as a designer and codify it, so an agent understands why something should be spaced the way it is, why the rhythm of a page matters, and why you shouldn’t put two primary buttons side by side.

Those are the pieces you pick up over years as a designer. The work was putting them into a form a model can follow, with the reasoning attached, so it can do the right thing every time instead of only when it happens to guess well.

04

Coverage

More components, more blocks.

Alongside the token and agent work, we expanded the system’s component and block coverage, including some of the more complex pieces their internal product needed.

  • 01Component coverageMore of the parts teams need, built on the new semantic tokens.
  • 02Block coverageLarger patterns assembled from those components, ready to drop into a page.
  • 03Complex tablesData tables built to handle the depth their products need.
  • 04Chat input for LLMsAn AI chat input for their internal product.
05

Outcome

Better design output across the company.

Over the last few months the work has helped improve the overall output of design at Applied Intuition, with a system that people and agents can both pick up and use to build something that feels like it came from Applied.

Let’s buildsomething withgood bones.

bones.ianc@gmail.com
© 2026 Ian BoneDrawn in BONES, this site’s own design system