Client
B2B SaaS vendor
Organization names are withheld due to confidentiality, but the industry and scale reflect a real project.
Industry / Theme
Product UI
Tags indicate the project’s primary theme and technology areas.
Period
2025
Years of the principal phases from kick-off through production launch. Operations support may still be ongoing.
Scope
Design system / UI component design / AI generation pipeline / engineering integration
We worked end-to-end across requirements, design, implementation, and post-launch operations.
01 — BACKGROUND
Background
A SaaS with 100+ features ships new UIs every quarter. With a small design team, both consistency and speed had hit a ceiling — the classic “the more screens we ship, the less consistent the design becomes” problem at scale.
02 — APPROACH
Approach
We taught the AI the design tokens (color, spacing, typography) and the pattern library, automating first drafts at the component and section level. Designers shifted to final polish and exception design. AI as “the partner that produces the first draft of a component,” humans as “the final line on quality.”
- Restructured Figma variables, components, and patterns into a form AI can consume
- AI generated initial wireframes for new screens; designers reviewed and adjusted
- Updating design tokens automatically propagates to generated work — a reproducible foundation
03 — RESULTS
Results
×3.5
UI delivery speed
(vs. baseline)
(vs. baseline)
- New-feature UIs shipped 3.5× faster
- Consistency score (internal metric) held steady; per-screen tweak time dropped sharply
- Designers reclaimed 20 hours/week for creative judgment work