You are no longer the user. You are the principal.

Most designers are still answering the same question they've been answering for thirty years.

How does the user do this?

Google I/O just changed the question.

Gemini Spark, Android Halo, WebMCP. Every announcement pointed toward the same thing. People stop operating software. Agents do it for them. The user promoted themselves to principal. They set intent. The agent executes.

This breaks the entire frame of modern UX. You cannot run a usability test on an agent. You cannot map a user journey when the journey runs while the user sleeps. You cannot design error states when the error is the agent misunderstanding the intent.

What you can design: the rules. What the agent is allowed to do. What it must ask permission for. How it communicates uncertainty. How it fails safely.

That is not a screen design problem. It is a behavioral contract problem.

Most design teams are not equipped for this yet. Most won't notice until the work is already being done by someone else.

What is one thing your product currently asks users to do manually that an agent should be handling?