Technology

How Figma Is Teaching AI to Stick to the Design Rules

Martin HollowayPublished 3w ago3 min readBased on 9 sources
Reading level
How Figma Is Teaching AI to Stick to the Design Rules
Photo by Apunto Group Agencia de publicidad on Pexels

Figma announced new tools at its first Config event in India to set how AI helpers work on its platform. The updates aim to keep AI output inside brand rules, let helpers do more on the main work canvas, and connect Figma to other AI tools. TechCrunch

The main release is Figma Specs, a simple text editor for writing down a brand's design rules for colors, fonts and parts. Think of it as a rulebook. A spec can be its own file or sit next to designs on the canvas. Users can link to canvas pieces from a spec and place FigJam diagrams, code layers and designs inside it.

Teams can put design libraries, code libraries and guidelines into Specs to keep context in one place. Helpers then check that central source to follow the rules. The stated goal is for Figma's own helpers and outside helpers to stick closer to brand rules when making designs with AI.

Figma also opened its helper on the canvas to more users. The helper in Figma Design can make and remix designs, do routine chores and give design feedback. Earlier work described the helper making design layers to show intent across flows, states, words and layout, with skills, or extra instructions, used to guide helpers with context.

Code layers are now in public beta, after a private beta in June. Code layers put working code into canvas designs for prototypes that act more like the real thing. Users can add one from the toolbar, make one from an existing frame, or ask the helper to make one. Users can ask the helper to turn a design into code or change part of the code to try a new version.

The other part is links to other tools. Figma now supports over 20 new tools through its MCP, a standard link that lets AI tools work with Figma, including Open Code, Dia, Grok Bot and Sarvam. That builds on a canvas that Figma says AI helpers can design directly on, an approach it laid out in March when it opened the canvas to helpers.

Figma is also joining Payload with its Figma Make product for making changing websites and managing that content directly. Figma bought Payload, an open-source system that stores website words and images apart from page layout, last year. Figma said Payload downloads are up sixfold in the last 12 months.

The venue reflects use. Figma said India is its biggest market by users outside the United States. It said its platform is used with 40% of India's top 100 companies listed on the Bombay Stock Exchange.

Co-founder and CEO Dylan Field has described this phase as "AI has lowered the floor but has not raised the ceiling, and that designers, creatives and builders will raise the ceiling." The helper is leaving beta with faster responses and the ability to search for Figma files in chat, while shader creation in Figma Design stays in open beta. Config 2026 launches included code layers, Figma Motion, shaders, generative plugins and Weave tools on the canvas.

The broader context here is making versus controlling. Many design tools can now make screens fast. The common failure is neat layouts that do not match saved settings, reusable parts and building limits. Specs keeps the rules where work happens in a form machines can read, not in separate notes helpers skip.

In my view, two practical tests matter. First, whether Specs stay current when code libraries change. Second, whether more MCP links allow steady trips between Figma and coding tools without losing layout. The hopeful case is simple. A shared rulebook plus working code and a content path through Make shortens the path from idea to working prototype with real content.