# What is agentic UX, and why does it matter?

Agentic UX is design for software that acts on a person's behalf. It answers 4 questions a chat box cannot. What will the agent do? When does it ask me? How do I follow what it is doing? How do I undo it? Get those right and people hand over work. Get them wrong and they watch every step.

By Surinder Thakur, AI Product Designer and Builder. Published 17 Nov 2026. Topic: [Getting your team to use it](https://www.surinder.design/work?type=getting-your-team-to-use-it).

## Why is a chat box not enough?

A chat box is good for asking. It is poor for watching, checking and undoing work.

Once software acts, people need more. A plan they can stop, progress they can follow, an undo they can find.

## What are the core patterns?

A plan shown before the work. Clear limits on what it does alone. Approvals only where a mistake is costly.

Progress in plain words, a readable log, and an undo window for actions that can be reversed.

## Who should design it?

Someone who understands both the work and the agent's limits. The hard decisions are about risk, not pixels.

Which steps run alone, where a person decides, what the approver sees. Those choices decide whether the tool gets used.

## How do you know it works?

People stop checking every step. They approve the right things, decline the odd ones, and use the agent weekly.

If they hover over it all day, the design has not earned their trust yet.

## In 1 picture

Software that acts needs design for watching, checking and undoing.

A chat box:
- Good for asking
- Work hidden in replies
- No clear undo

Agentic UX:
- A plan you can stop
- Progress you can follow
- An undo you can find

## Next questions

- [How do you show people what an AI agent is doing while it works?](https://www.surinder.design/work/how-to-show-what-an-ai-agent-is-doing): Show 4 things. The plan before it starts, so people can stop it early.
- [How do you design an AI agent approval screen people actually read?](https://www.surinder.design/work/how-to-design-an-ai-agent-approval-screen): Show 4 things on 1 screen: the action, the exact amount, the reason, and the rule that stopped it.
- [What should an AI agent be allowed to do without asking?](https://www.surinder.design/work/what-should-an-ai-agent-do-without-asking): Anything a person can check and undo in time. Reading, sorting and drafting need no approval, because nothing leaves the building.
- [Is my design system ready for AI agents to use?](https://www.surinder.design/work/is-my-design-system-ready-for-ai-agents): It is ready when an agent can read it as files, not pictures.

Page: https://www.surinder.design/work/what-is-agentic-ux
