UX Collective: Rethinking Figma in an AI world
Figma occupies a rare position in the design tool landscape: it became the default workspace not just for visual design but for the entire cross-functional conversation around product decisions. That position is now under structural pressure, and Darren Yeo’s June 2026 article on UX Collective examines what that pressure looks like and what Figma is doing about it.
The article is not a product review or a prediction that Figma will fail. It is a careful analysis of a business model stress test, written from the perspective of someone who has used Figma extensively and watched Config 2026 with attention to what the announcements were designed to protect.
The threatened business model
Figma’s revenue comes largely from seat-based licensing—paying for every person who accesses and works in design files. That model made sense when design files were the primary artefact that engineers, PMs, and stakeholders needed to reference. As AI tools allow engineers to generate and inspect UI directly inside coding environments, the passive stakeholder seat becomes less necessary. Yeo argues that this is not a distant hypothetical; it is already visible in teams where engineers are iterating on UI in Cursor or Claude Code without opening a Figma file at all.
Config 2026 as defensive expansion
Yeo reads Figma’s Config 2026 announcements—code layers, motion, shaders, agentic workflows, MCP integration—as the company expanding its canvas rather than protecting the canvas as it currently exists. Instead of defending the idea that design must happen in Figma, the company is trying to make Figma the place where design context lives, even when the actual generation happens elsewhere. The MCP server, which allows external agents to access design tokens, components, and layout context, is particularly significant in this reading: it lets Figma become a source of truth for AI outputs rather than a production environment for human outputs.
Anthropic’s competitive pressure
The article gives meaningful attention to Claude Code and what Yeo calls “Claude Design” tools normalizing workflows that bypass design files entirely. When a team can prompt production-quality UI from a description and ship it without a Figma step, the question of what role Figma plays in that process becomes urgent. Yeo’s framing: the real question is whether design intent becomes portable across tools or remains locked to a file format.
What this means for practitioners
For designers, the article raises a practical question: what is your advantage in a workflow where code is increasingly generated rather than specified? Yeo suggests the answer is not learning to write code but rather developing the ability to articulate design intent in forms that are useful to AI agents—whether through well-maintained design systems, annotated token structures, or the kind of context that makes AI outputs converge on correct results rather than approximate ones.
The analysis is honest about uncertainty. Yeo does not claim to know how this resolves, but the framing is useful: teams that treat Figma as the place where design intent is encoded will fare better than teams that treat it as the place where screens are produced.