Harvey AI Overhauls Design System as Legal Tech Unicorn Scales Operations

Harvey AI, the legal-focused AI company valued at $1.5 billion following its June 2024 funding round, has completed a ground-up rebuild of its design system to address what the company called "chaotic" infrastructure that couldn't keep pace with product expansion.

The overhaul tackles a common scaling problem in fast-growing startups: design and engineering teams speaking different languages. Harvey's previous setup mixed legacy components, Shadcn components, and custom one-off implementations across its Figma designs and React codebase, creating translation errors that slowed development.

What Actually Changed

The company moved from appearance-based naming ("neutral-400") to intent-based semantic tokens ("foreground-base"). This sounds like inside baseball, but it matters for anyone watching Harvey's expansion trajectory. Semantic tokens mean the company can now ship new themes, accessibility modes, or product brands without rebuilding everything from scratch.

Harvey built a GitHub Action that syncs Figma variables directly to code. Designers update tokens in Figma, trigger a workflow, and a pull request opens automatically. No more manual handoffs or interpretation errors between design and engineering.

The migration strategy reveals how seriously Harvey is treating operational efficiency. Rather than a hard cutoff, the company implemented graduated enforcement: linter warnings for primitive token usage now, warnings for legacy tokens at 80% adoption, and pre-commit errors at 95%.

Why This Matters for Harvey's Roadmap

The timing aligns with Harvey's push into new platforms. The company explicitly mentions mobile and Microsoft Office add-ins as immediate priorities, requiring components that "feel native to each platform, while still unmistakably Harvey."

For a company serving elite law firms—where trust and precision are non-negotiable—this infrastructure work is foundational. Harvey's design philosophy emphasizes transparency and domain awareness, principles that require consistent execution across every surface.

The $100 million raised in June 2024 appears to be funding exactly this kind of operational maturity. Harvey is hiring for design system roles across both Design and Engineering teams, signaling continued investment in the foundation rather than just feature velocity.

Whether this positions Harvey to widen its lead in legal AI depends on execution. But the company now has infrastructure that can flex without requiring another major overhaul—a prerequisite for the multi-platform expansion they're telegraphing.