What if you could describe a design system and watch it build itself?

If you've worked in product design for any length of time, you already know this: a proper design system with semantic tokens, multi-mode variables, and documented components is the foundation of scalable work. But building one from scratch in Figma takes 2 to 4 weeks of careful, methodical effort. For freelancers balancing multiple clients and small agencies running lean, that's time they rarely have.

So what ends up happening? Most designers either skip the system entirely (and pay for it later in inconsistency and mounting tech debt) or buy a premium UI kit that gets them maybe 60% of the way there, then spend days customizing it to match a client's brand. Neither option feels great.

I spent several weeks mapping every tool in the AI + Figma space, and what I found was a clear, open gap. Figma Make generates code-first prototypes (limited to 3 frames). Relume builds sitemaps and wireframes. Lovable and Bolt skip Figma entirely. Nobody was building conversational AI that writes native design system components directly to the Figma canvas.

Meanwhile, Figma's own data showed that 30% of their highest-value enterprise customers were already using AI-assisted design on a weekly basis. The demand was clearly there. The solution just didn't exist yet.

Intent Studio is a Figma plugin with a conversational AI interface. You open it, describe what you need ("Build a fintech dashboard design system with accessible slate-neutral variables and a Tailwind-aligned spacing scale"), and the AI orchestrates the entire creation process in real time, writing directly to your Figma file.

No code generation. No HTML conversion. No exporting and importing. Real Figma variables, real components, real auto-layout. All built through a steady, back-and-forth conversation.

The core technical challenge here was reliability. Generative AI is inherently unpredictable, but design systems require deterministic precision. Every token needs to resolve, every alias needs to connect, every component needs proper structure. There's no room for "close enough."

I solved this by designing a structured command protocol: 45+ discrete JSON command types that the AI generates and the plugin executes, rather than letting the AI write arbitrary code. This keeps things predictable, secure, and consistent every time.

The business model lives or dies on unit economics. "Unlimited design systems" is a compelling pitch, but only if each generation doesn't cost more than the subscription brings in. I approached cost optimization as a product design problem, not something to figure out later.

Three strategies compounded together to make this work: structural section kits (don't regenerate what you can hydrate), intelligent model routing (Opus for thinking, Haiku for doing), and output optimization (compressed command batching). The result is sustainable margins that make "unlimited" pricing honest and viable for a solo-founded business.

Before investing in go-to-market, I ran extensive local testing through Figma's Desktop Bridge to answer the core question: can conversational AI produce design systems that a professional designer would actually trust and use in their work?

As a solo founder, the GTM strategy needed to be focused and realistic. I'm prioritizing channels that compound over time: Figma Community discovery (the primary surface where designers find new tools), build-in-public content showing real generations, and targeted outreach through 12+ years of agency relationships.

The launch plan sequences carefully: beta testing, then community launch, then Product Hunt. The broader business model follows a "land and expand" approach where the $19–29/mo design tool builds trust and familiarity, which eventually opens the door to a $79–149/mo CMS pipeline (Track 1) that turns design systems into live websites.

Intent Studio is, at its core, a bet on a simple principle: the most powerful design tools will be the ones that let you describe what you want, and then get out of the way. For a lot of execution tasks, the AI genuinely produces better, more consistent results when the human focuses on intent and direction rather than pixel-by-pixel construction.

The hardest problems weren't the AI itself. They were the systems thinking: designing a token architecture that scales gracefully, a command protocol that stays deterministic, and a cost model that makes "unlimited" an honest promise. Those are product design challenges as much as engineering ones, and they're the parts I'm most proud of working through.

Building a product solo forces a certain clarity about what matters. Every feature has to justify itself against one question: does this help someone reach their first "wow" moment faster? That discipline shaped everything from the 5-pass architecture to the decision to hold off on CMS features until the design tool has clearly found its footing.