Unifying three design systems into one AI-ready foundation
When I joined as Director of Design Systems & Foundations, I inherited three different design systems running in parallel — inconsistent UX, drifting visual design, broken components, and no shared direction. The most promising of the three was tightly coupled to a single UI platform, and none of it was built for the way software gets made now: there was no AI presence anywhere. This is the story of how I turned that fragmented inheritance into Summit — a single, platform-agnostic, AI-friendly design system — and the strategy that made it adoptable across the company.
The role
Director of Design Systems & Foundations · Constant Contact. Owner of Summit, a single, platform-agnostic, AI-friendly design system spanning every product surface.
The goal
Give every Constant Contact product one coherent, trustworthy foundation — one that works regardless of UI platform and that both humans and AI can read, reason about, and build with. Consistency was the surface goal; leverage was the real one. A change made once in the system should ripple everywhere, instead of being re-litigated team by team.
The problem
Three systems meant three sources of truth, and none of them was complete. The same component looked and behaved differently depending on which system a team pulled from. There was no roadmap, no ownership model, and no shared direction. The leading system was locked to a specific UI platform, making it unusable for teams building anywhere else. And documentation wasn’t legible to AI — no AI-powered workflows, and no components designed for AI-driven experiences. Every product-wide design change turned into a manual, team-by-team slog.
“There were three systems and no one could tell me which one was the right one to build on. So I’d just pick one and hope.”
— Product Designer
Business impact
Fragmentation is expensive in ways that don’t show up on a single line item. Engineering time was lost rebuilding components that already existed elsewhere; design time was lost reconciling three visual languages; and the inconsistency eroded the polished, trustworthy feel customers expect from the product.
A guiding principle
Build a system that’s legible to both people and machines. Every decision — how tokens were structured, how docs were written, how components were packaged — was made so a human designer and an AI agent could understand it, trust it, and build with it without a human translator in the loop.
Our approach
I started with strategy, not components. Before refactoring a single token, I defined what the system was for, who owned it, how teams would migrate onto it, and how we’d sunset what we were replacing. That strategy became the spine every workstream below hangs from.
Refactored the token library into a two-tiered semantic model
I rebuilt the unfinished token library as a two-tiered semantic library — a primitive layer of raw values and a semantic layer that expresses intent (purpose, state, hierarchy). Semantic tokens are what make the system themeable, platform-agnostic, and legible: a token named for what it means rather than what it is can be reasoned about by a designer, a developer, and an AI tool alike.
Rebranded the design system: Rise → Summit
I rebranded the design system from Rise to Summit — not just a name, but a reset. The rebrand signaled a clean break from the fragmented past and gave teams a single banner to rally behind.
Made the system AI-native, end to end
AI-legibility wasn’t a feature bolted on — it’s the thing that makes the whole system compound. AI generates all of our component documentation, including the changelog, so docs stay accurate as the system evolves. AI posts design-system changes into our Slack channels, so every update reaches the teams who consume Summit automatically. AI generates prototypes that use Summit as its guidelines, producing production-ready UI code rather than throwaway mockups. And I shipped components built for the AI era — the AI Mark, our branded AI logo, and a left-rail chat experience packaged as first-class system components.
Decommissioned legacy systems and migrated products
I’m decommissioning two legacy design systems and migrating their products onto Summit — collapsing three sources of truth back into one.
Supported a company-wide brand refresh through token sharing
I supported a broader brand refresh — extending all the way to the front-of-site — by sharing tokens across surfaces. Because brand decisions live in the semantic token layer, a single change propagates from the marketing site through the product, keeping the whole experience in sync.
Results & business impact
Summit replaces three fragmented systems with one coherent, AI-ready foundation: one source of truth, semantic tokens that scale brand changes everywhere at once, AI-generated documentation and changelogs, and AI-specific components shipping across products. The result is consistency customers can feel and leverage the whole organization can build on.