90% of Your AI Agent's Design Process Is Dead

The traditional design process built around Figma mockups, handoff documents, and sequential engineering is becoming obsolete as AI agents dramatically reduce the cost of building the wrong thing. Jenny Wen, head of design at Anthropic, notes that designers now spend 30-40% of their time on prototyping instead of 60-70%. The new workflow starts with defining actors (users with specific goals), identifying where views split, and setting constraints before touching any code. A PRD is generated via an AI interview prompt, then converted into an architecture file listing pages, modals, and user flows. An AI agent then builds a Next.js frontend prototype with mock data for client approval. Once approved, the agent generates an API spec and Supabase schema, uses the Supabase MCP to automate database setup, and connects the frontend directly. A separate backend layer for payments, notifications, and analytics is handled by a cloud agent. The key insight is that planning (actors, flows, constraints) remains essential, but the translation layer between design and code has been eliminated.