The Agentic UX Playbook
Agents aren't screens. They are work happening over time.
Six moments decide whether people can trust that work. This playbook shows how to design each one, how it looks in practice, and how sure we are.
What's inside the PDF
The website explains it. The guides help you ship it.
6guides
90pages
15per moment
The entries
- 01 OnboardingBefore it starts How should someone first meet an AI agent? Free entry · Also in the PDF
- 02 WaitingWhile it works Is the agent working, or is it stuck? Free entry · Also in the PDF
- 03 DiscoveringWhile it works It found something you did not ask about. Now what? Free entry · Also in the PDF
- 04 ApprovingWhen it needs the person When should an AI agent stop and ask? Free entry · Also in the PDF
- 05 InterruptingWhen it needs the person What does Stop actually do? Free entry · Also in the PDF
- 06 RecoveringAfter it finishes It failed halfway. Now what? Free entry · Also in the PDF
This is a working playbook. New entries are added when they are finished. Prefer cards? See the Agentic UX Toolkit →
Three questions for any agent
This is our own way of looking at it, not an established framework. The more an agent can do alone, the more the other two matter.
What can it do alone?From suggesting an answer to finishing a multi-step job by itself.
Can I see what it did?Enough to understand and check it, without the technical noise.
Can I step in?Approve, stop, change direction or undo, at the moments that matter.
All six guides, ₹699
Frame-by-frame examples, copy you can use, a checklist for design reviews and a build prompt for each moment.