Work Practice About CV
← All work

The Design Uplift Initiative

Our Storybook library drifted from the Figma design system. Developers faced a daily choice between stale components and rebuilding from scratch. The industry calls this design system drift — and 41% of design systems launched in the last two years have already been abandoned to it. Not because the components were bad. Because governance never followed the build.

Inspira Financial Design Systems Manager 2026 — Phase 1 in progress

The problem nobody budgets for

A design system is not a component library. A component library is what you get when you build one and walk away — and two years later, 41% of them have been abandoned.

That figure comes from Knapsack’s 2024 Design Systems Maturity Report, and the reason it’s so high isn’t a tooling problem. The components were fine. What failed was governance: no named owner, no contribution path, no versioning. The system drifted from the design source, developers stopped trusting it, and the whole investment quietly evaporated.

What I’m doing about it

Migrating a Figma design system into a current Storybook library — AI-accelerated — while building the governance layer first rather than bolting it on afterward.

The AI part is the easy part, and I want to be precise about that, because most people have it backwards. An agent reading Figma can generate a component and its stories in minutes. What it cannot do is decide what “primary” means, notice that the Figma layer names don’t match the code conventions, catch the edge cases a static design file can’t express, or own the thing six months from now.

So the workflow puts a human at every judgment point and lets the AI do the typing:

  1. Audit — the agent reads the component: tokens, variants, states
  2. Generate — component file plus stories
  3. Review — I check props, accessibility, and the edge cases Figma can’t show
  4. PR — an engineer approves. They read; they don’t rebuild.
  5. Merge — tracker updates, component goes live

The detail that decides the outcome

Naming alignment.

When Figma layer and variant names don’t match code conventions, generated code drifts immediately — and so does hand-built code, which is the tell that this was never really an AI problem. “Primary” in Figma has to become primary in code. Not main. Not type1.

Fixing that at the source, once, before generating a single component, is dramatically cheaper than correcting it downstream forever. It’s also the least glamorous week of the project, which is exactly why teams skip it.

What I’ll publish

I’m writing this one as it happens, phase by phase — including what the AI gets wrong. That’s the part nobody publishes, and it’s the part I’d most want to read.

Detailed strategy, RACI, and phase gates available on request.