Figma's AI Agents Can Now Design For You. Most Teams Aren't Ready.

Yesterday, Figma announced that AI agents can now write directly to your Figma files. Not generate images and paste them in. Not suggest layout options for a human to apply. Actually write to the canvas, creating and modifying real design assets using your components, variables, and tokens. Through the Figma MCP server, tools like GitHub Copilot can now push rendered UI into your files and pull design context back into code.

This is a different category of shift from everything that's come before.

I've been building digital products with AI since 2018. Back then it was classification models in banking flows, anomaly detection in travel, content recommendation in super-apps. None of it was 'the AI is designing the product.' It was always 'the AI is changing what the product needs to do, and we're designing around that.' That distinction mattered. The human was still making.

Now the tool is making. And the gap between 'the AI helped me design this' and 'the AI designed this' just got very small.

What strikes me isn't fear about replacement. What strikes me is that most design teams are completely unprepared for the governance question this raises. When an agent reads your design system, interprets your components, and creates production-ready frames, what's your review process? Who owns quality? How do you catch the agent building technically correct but contextually wrong output?

I ran a neobank MVP from zero to $7M in eight weeks. Thirty thousand card orders. That kind of speed is only possible when your system is tight. Every component named consistently, every token scoped correctly, every pattern documented with intent, not just implementation. At that pace, a sloppy system doesn't slow you down gradually. It kills you suddenly. A designer making mistakes at that speed is bad. An agent making mistakes at that speed is a different problem entirely, because it's consistent, fast, and hard to spot until it's in production.

At RUONALIM, this is how we approach every client engagement in Dubai. System hygiene first. Because when agents start building on top of your components, the quality of your foundation becomes the quality of your output.

The teams that will actually benefit from agentic Figma are the teams who already have their houses in order. Clean naming conventions. Documented decision rationale, not just spec'd outputs. Components built with intent rather than just built. That's always been good practice. It's now table stakes. If your design system is a graveyard of half-connected components and renamed-once-then-forgotten tokens, an agent running on top of it will produce beautifully broken output at scale.

The design leader's job just became more editorial, more architectural, and more political, in the best sense of that word. You're setting standards that AI will execute against. That's a different skill set from the one most IC designers are practising.

I'm not saying this is bad. I think it's clarifying. The designers who thrive in this will be the ones who always knew that making wasn't the point. Understanding what to make, and why, and for whom. That's the job. AI executing on a well-reasoned design brief is a force multiplier. AI executing on a vague one will produce beautiful noise.