AI-powered UI generation often suffers from an "uncanny valley" of design where layouts look professional, but components are generic and the data is purely fictional. Figma is addressing this lack of relevance by introducing Make kits and Make attachments, two features designed to anchor AI outputs in structural and data-driven reality.

Make kits allow design teams to sync their specific design systems—via npm packages or Figma libraries—directly into the generation process. This ensures that the "first draft" created by the AI isn't just a pretty picture, but a functional layout using the exact components and styles found in the production codebase. By teaching the AI the specific rules of a company's design language, teams can bypass the tedious manual cleanup usually required to align AI mocks with internal standards.

Complementing this is Make attachments, which provides project-specific context. Users can now upload various file types, including CSVs for data or PDFs for legal requirements, forcing the AI to work with real-world constraints. Instead of generating idealized versions of a screen, the AI can now account for long text strings, specific data fields, and brand guidelines, moving the needle from visual approximation to a grounded, ship-ready prototype.

AI-powered UI generation often suffers from an "uncanny valley" of design where layouts look professional, but components are generic and the data is purely fictional. Figma is addressing this lack of relevance by introducing Make kits and Make attachments, two features designed to anchor AI outputs in structural and data-driven reality.

Make kits allow design teams to sync their specific design systems—via npm packages or Figma libraries—directly into the generation process. This ensures that the "first draft" created by the AI isn't just a pretty picture, but a functional layout using the exact components and styles found in the production codebase. By teaching the AI the specific rules of a company's design language, teams can bypass the tedious manual cleanup usually required to align AI mocks with internal standards.

Complementing this is Make attachments, which provides project-specific context. Users can now upload various file types, including CSVs for data or PDFs for legal requirements, forcing the AI to work with real-world constraints. Instead of generating idealized versions of a screen, the AI can now account for long text strings, specific data fields, and brand guidelines, moving the needle from visual approximation to a grounded, ship-ready prototype.