Before the agent acts
Start with the goal, narrow scope, preview the plan, and decide what the agent is allowed to do.
The pattern language for AI that acts on your behalf
Lifecycle patterns for outcome-oriented agents—align before action, show work in progress, make recovery explicit. Based on established patterns from apps in production.
Lifecycle framework
Agentic UX groups patterns by when they surface in a supervised run—not by UI widget or model capability. Before work starts, interfaces align before action: scope, plan, and permission. While the agent executes, they show work in progress, rationale, and uncertainty. After it finishes, they audit outcomes, make recovery explicit, and support handoff. Walk the three phases in order when auditing a flow, or jump to the stage where your product feels weakest. Read the framework article for the full thesis and pattern matrix.
While the agent works
Show work in progress, explain choices, assemble fit-for-task UI, and preview work in a safe sandbox.
After the agent acts
Audit what changed, make recovery explicit, escalate when stakes rise, and brief users when they return.
Agentic UX, made practical
What this is
An interactive reference for supervised delegation: lifecycle patterns with limitations, Do/Avoid guidance, interactive mockups, and annotated production screenshots with critique.
Who it's for
Designers and builders shipping apps where agents run work on behalf of users. Use it when you need trust, visibility, and explicit recovery—not chat-only interfaces.