Agentic Experience Patterns

How humans hand work to agents — scoping, briefing, and transferring intent.

How agents return work to humans — summaries, diffs, and receipts.

Logs, traces, and audit trails that make agent behavior inspectable.

How agents expose their plan before acting, and how humans shape it.

How interfaces communicate what an agent is doing and how far along it is.

Gates where an agent must ask before acting — the design of consent.

How humans stop, pause, or redirect an agent mid-task.

Stop, retry, undo, rollback, escalation, and scope reduction after failure.

How humans inspect and evaluate agent output before accepting it.

How agents are named, attributed, and distinguished from humans.

How agents are granted, scoped, and revoked access to tools and data.

How context is gathered, displayed, scoped, and shaped as a design material.

What agents remember across sessions, and how users see and edit it.

Interaction between multiple agents — coordination, negotiation, delegation chains.

Agents working unattended — visibility, notification, and re-engagement.

Interfaces generated or composed by agents at runtime.