Mike Hakobyan
← Back to work

Making a 150+ component design system AI-ready

I rebuilt EatApp's design system as tokens, documented it for designers, developers and AI, and connected it to Claude through Figma MCP.

Some details are generalized or blurred to respect client confidentiality.

Role
Senior Product Designer, leading the tokenization and the AI workflow
Team
Small design team, working with product and engineering
Timeline
2026
Tools
Claude, Claude Code, Claude Design, Figma MCP
150+components moved into a tokenized system
2 layersprimitives + semantic tokens, in light and dark
10x+faster on token setup, docs and early explorationsMy estimate

The system worked, but it didn't scale

01

Inconsistent components

Working files had many near-duplicate variants. The side navigation alone existed in several slightly different versions.

02

No real token layer

The colour guidance was a generic template palette that didn't match EatApp's green brand, and values were hardcoded across files.

03

Slow, repetitive work

One change meant editing many screens by hand.

04

AI didn't know our design language

AI tools produced generic UI that didn't look or behave like EatApp.

The real blocker for using AI wasn't the AI. It was that the system wasn't structured enough for AI to understand.
The working canvas before: template colour guidance mixed with loose button variants.
The side navigation, in multiple drifting variants.
The product as it looked (booking details replaced with placeholders).

Audit → Tokenize → Document → Connect → Generate → Review

Audit the system with AI

Using Claude with Figma MCP, I audited 150+ components for hardcoded values, duplicates and drifting variants. This gave us a clear map of what to fix before building anything new.

One of the audited working areas.

Tokenize

I rebuilt the foundations as two layers. Both work in Light and Dark mode, with radius and shadow scales on top. Now one change updates everywhere.

10x+ fasteron token setup, docs and early explorations (my estimate)
Layer 1

Primitives

Raw scales, like green-500 or grey-900.

ShadeNeutralPrimarySecondaryErrorWarning
Layer 2

Semantic tokens

Tokens that describe what a colour does.

SurfaceForegroundBrandStatusBorder
Primitives, light mode
Primitives, dark mode
Semantic tokens

Document the rules for humans and AI

I used Claude to generate the documentation, written so it works for three audiences. For AI, it all comes together in one file: DESIGN.md.

For Designers

Component docs with states, props and usage. Dropdowns, Pagination and Layout grids are examples.

For Developers

A token reference that maps every Figma token to its CSS variable and Tailwind class.

For AI

DESIGN.md: explicit do/don't rules that keep generated output on-system.

One file that teaches AI how to use our design system

The system itself was standardized with AI, following a structure Claude came up with. We captured that structure in DESIGN.md: which semantic token to use where, which radius goes on which component, and what never to do.

Then DESIGN.md guided AI on how to use the system, so the designs it generated used the correct components with the correct variables.

  1. 1AI standardizes the system
  2. 2Its structure becomes DESIGN.md
  3. 3DESIGN.md guides AI to generate on-system designs
Stack contextColour tokensSurface · Foreground · Brand · Border · StatusSpacingBorder radiusUsage patternsDo not
DESIGN.mdRead by Claude
# Design System > All tokens are semantic — use these, never raw hex values or primitive names. ## Colour tokens · Surface Surface/Sunken bg-surface-sunken → input fields, textarea, select backgrounds Surface/Selected bg-surface-selected → selected table row, active nav item ## Border radius rounded-md 8px buttons, inputs, chips rounded-lg 12px cards, panels, dropdowns rounded-3xl 24px dialogs, modals ## Do not - Use raw hex values - Use primitive names (Grey-800, Green-500) - Use arbitrary values (p-[20px], rounded-[8px]) - Use Error/Fill for alerts — use Error/Surface
Developer token reference
Radius rules for AI
Dropdown docs
Pagination docs

Connect AI to the real system

With Figma MCP, Claude worked directly with our Figma files and tokens, with DESIGN.md telling it how to use them, instead of guessing. That's the step that fixed the "generic AI output" problem.

Figma design system150+ components
Tokens + DESIGN.mdSemantic tokens, usage rules
Figma MCPDirect access to files
ClaudeWorks with the real system
On-system designsEatApp's design language
DesignSystem + workflowAI

Generate new designs from the system

We used Claude and Claude Design to produce complete explorations from prompts, built on EatApp's real components and tokens. Every generation was guided by DESIGN.md, so screens came out using the right components with the right variables.

5A

A modular navigation system

App shell, layered navigation, page structure, and "one shell, many pages," plus a before/after comparison and sidebar states.

ResultExploration
Generated before vs after summary
Exploration: generated sidebar states built on the new systemBefore: the side navigation in many drifting variantsBeforeExploration

Drag to compare the drifting sidebar variants with the generated sidebar states.

5B

A multi-step messaging campaign flow

Define → Audience → Template → Schedule. Redrawn here without any real data.

ResultExploration
5C

An in-app AI assistant concept

Language detection and switching, file upload, and guest message translation, plus the logic behind them.

ResultExploration

Language detection & switch

The guest's language is detected automatically and confirmed inline, asked explicitly when unclear, and can be changed at any time via an always-available switcher.

  1. 01Auto-detect & acknowledge
  2. 02Explicit ask (fallback)
  3. 03Always-available switch

The logic behind it

How the assistant decides a guest's language: from the guest's explicit choice down to their history, with settings each restaurant can configure.

Human review before handoff

AI produced the options, and we made the decisions. I stayed in the loop on two things:

UX judgment

I chose the right direction, fixed flows and made sure screens worked for real restaurant use.

Checking for hallucinations before handoff

I verified that every generated screen used real components and tokens, with no invented patterns, before it went to engineering.

Human decision
AI produced the options, and we made the decisions.

One system, from Figma to code to AI

  • 150+ components moved into a tokenized, documented system with light and dark modes.
  • One source of truth from Figma tokens to CSS variables to Tailwind classes.
  • 10x+ faster on token setup, docs and early explorations (my estimate).
  • AI output matched EatApp's design language because it was built on the real system instead of generic patterns.
  • Designers spent more time on product decisions and less on production upkeep.
150+components tokenized
1source of truth
2modes: light + dark
10x+faster setup, docs and explorations (my estimate)

Structure first, AI second

Structure first, AI second

AI is only as good as the system you give it. Tokens and rules did most of the heavy lifting.

Write docs for AI, too

Clear do/don't rules are the difference between on-brand output and generic output.

AI speeds up exploration, humans own the decisions

The value came from generating many options and judging them well.

A review step is non-negotiable

Checking for hallucinations before handoff makes AI output safe to ship.

What I can do for your team

  1. 01Audit your design system
  2. 02Tokenize it (primitives + semantic, light + dark)
  3. 03Document it for designers, developers and AI
  4. 04Connect it to AI through MCP
  5. 05Generate designs in your real design language
  6. 06Keep humans in the loop on decisions and quality

Want this for your team?

Visuals are blurred or simplified to respect confidentiality.