Designed to be read: making a system AI can actually use
summary
Meta describes rebuilding a legacy design system with Figma to make AI outputs actually usable, co‑developing a Figma Make integration and a Model Context Protocol (MCP) that supplies color tokens, components and docs at request time. The team sets an 80% high‑accuracy bar, reaches about 87% in a pilot, produces pixel‑exact components, and publishes a three‑step playbook: audit context, embrace iterative feedback loops, and prepare for agents.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 2:43AI design tools fail at scale
- 4:29AI design failures root causes
- 5:51Quality with golden sets and iteration
- 7:58Model Context Protocol with Figma
- 9:23AI-driven design layered stack
- 11:55Prepare for AI agents
- 13:20AI slop to quality at scale
- 14:23Multimodal design workflows future
related talks
12mThe future of design systems and AI
24mHow Agentic AI Empowers Creative Thinking
14mEmpowering designers: AI, design systems, and quality
33mBuilding an Agentic Design System
56mEnabling teams with AI
18mAI for UI
31mDesign to Code in 2026: Context-First Development with Figma MCP
17mDesign systems in an AI first ecosystem