The Canvas Is Becoming the Agent Workspace
The next useful AI product surface is not a detached chat box. It is the shared workspace where teams already shape the thing they are building.
The strongest June 25 signal from Digg Tech was Figma announcing Code Layers, Figma Motion, Weave generation, and an integrated agent for plugins and shaders. The important move is not that design tools added another AI button. It is that code, motion, model workflows, and custom tools are being pulled into the same canvas where product teams already collaborate. The workbench is becoming the interface for generation.
Product builders should read this as a shift in agent UX. If AI work happens outside the artifact, users lose context, review becomes awkward, and handoff costs rise. When agents operate inside the canvas, doc, board, repo, or issue where decisions already live, the output can be inspected, edited, and reused by the team. That is the difference between a demo and a working product loop.
Audit one AI feature and ask where the agent should live. Trigger: the user is already editing a design, spec, dataset, issue, or workflow. Context: the current artifact, surrounding comments, constraints, and history. Tools: generation, code, data transforms, motion, review, and export. Verifier: the artifact itself remains editable and understandable. Budget: scoped permissions, bounded runtime, and visible changes. Stop condition: the user can review and continue the work without translating it from a separate chat.
Put it to work 12 minutes
Evaluated context pack
Take one feature in your product that currently opens an AI side panel. Redesign it so the AI acts directly on the working artifact, leaves a visible trace, and lets the user accept, edit, or reject the result in place.
Your turn
Use the task above. Record the result and anything you still need to check.
Your draft stays in this browser. No account needed.