The tool sequence designers actually used, by discipline. See how the stack shifts as AI takes more of the execution.
PM and designer co-write a brief. Agents generate flows, states, and code. Human edits the parts users touch most.
59 synthesized monthsin the data layer. Stage breakdowns (Starter / Scaler / Titan) are available for 2026 only — earlier months show under the All segment but won’t appear under stage filters until the design-context pipeline runs further back.
Prompt-to-prototype-to-production is one motion. Product designers direct agents, write specs in natural language, and review generated flows against a taste rubric.
Product designers ship code paths directly. The PRD is a prompt. Review moves from pixel critique to judgment on whether the thing should exist.
Product designers in Q3 2026 spent the first half of the week in Figma designing agent-facing surfaces — the review screens, the plan-with-agent flows, the interrupt states. The second half was in v0 or Cursor, testing whether those flows held up when a real model was behind them. Prototyping in code became the norm for anything involving AI behavior; static mocks couldn't capture what happened when an agent went sideways.
Product designers in Q3 2026 open with a Claude brief — problem framing before any pixels. v0 or Figma AI generates a first screen in minutes; the real work is the edit pass: tightening hierarchy, pressure-testing the flow, making sure components are structured for agent-readability. Code-adjacent work is no longer optional — most mid-level product designers are touching Cursor at least once a week to verify that what they designed actually builds.
Product designers start in Figma to map flows and set the interaction intent, then hand the component spec to v0 to generate the React scaffold. Cursor with MCP pulls the live design tokens so AI-generated code doesn't drift from the system. The back half of the week is QA — running Playwright tests on shipped interactions and writing the rationale that becomes the next prompt rule.
Product designers open a sprint by pressure-testing the problem in ChatGPT or Claude, then move into Figma for structure. Execution work — initial layouts, component variants, first-pass screens — increasingly goes through v0 or Cursor, with the designer editing outputs rather than building from scratch. The hardest part of the week isn't making; it's deciding which AI-generated pattern actually solves the problem and which just looks like it does.
Product designers started flows with Google Stitch or v0 — generating five-screen canvases from a single prompt to pressure-test navigation logic before touching Figma. From there, they refined interaction details inside Figma, used AI plugins to generate copy and component variants, then pushed to Maze or Lyssna for rapid usability validation. The blank-canvas starting point is effectively dead; the job is now editing and directing generated flows rather than drawing them.