Agent Experiences

The agentic experience designer

What the role involves, the skills it takes, and how it differs from the design jobs it grew out of.

In one paragraph

An agentic experience designer works out how people hand tasks to AI agents (software that can carry out tasks for you as well as answer questions) while staying in charge of the outcome. A product designer shapes what someone does themselves, and this role shapes what happens when an AI model does it for them.

Why the role appeared

For most of the history of software, what you could see on screen and what the software could do were the same thing. If there was no delete button, nothing got deleted. The buttons and controls a designer put on screen marked the edge of what could happen, which is why so much design practice is about making the next step obvious and the current state visible.

Agents break that link. What an agent can do depends on the tools it has been given (the connections that let it search, send email, or change records in other software), and that is almost always more than anything shown on screen. The screen no longer tells users what is possible, and they are no longer the ones doing the steps. Feedback that used to arrive during an action now arrives after it, and mistakes a greyed-out button used to prevent now have to be caught, explained, and undone.

Someone has to make those decisions on purpose. When nobody owns them, they get made by default, usually by whichever engineer connected the agent to its tools. The result is an app that asks people to trust something they can’t see, predict, or stop.

What the work consists of

Deciding what the agent may do unsupervised

Sorting actions by what they could cost and whether they can be undone, then deciding which ones go ahead quietly, which wait for approval, and which are never handed to the agent at all. Reading a page and spending money shouldn’t need the same level of permission, and where to draw the line between them is a design decision, not an engineering one.

Showing what the agent plans to do before it does it

Describing the next step in words the user can judge, early enough that stopping it is still cheap. This is harder than it sounds, because a plan needs enough detail to judge and little enough that someone will read it.

Designing interruption and correction

A stop button that works partway through a task, a way to change direction without starting over, and a way back from a wrong action. Undo is a design problem here as much as an engineering one. Some agent actions reach systems you don’t control and can’t be reversed, and the interface has to say plainly which ones.

Getting trust to the right level

Helping people form an accurate picture of what the agent can be relied on to do. It can go wrong in both directions. Trust it too much and important actions go through without anyone checking them. Trust it too little and people end up supervising everything, so handing over the work saves them nothing.

Designing the report afterwards

What the agent says it did, in how much detail, and how someone checks it later. When the agent works with nobody watching, this report is the whole experience.

How it differs from adjacent roles

RoleThe main questionWhat they mostly produce
Product designerCan the user complete this task?Flows and screens
Conversation designerDid we understand each other?Dialogue, and ways to recover from misunderstandings
Agentic experience designerDoes the user understand what is about to happen, and can they stop it?Rules for what needs permission, when to hand over to a human, and reports of work done

Here’s a rough test. If you can list every screen a user can reach by clicking, you are doing product design. If you can list every turn the conversation might take, you are doing conversation design. If you can list neither, because the system decides its own steps to meet a request that was only partly spelled out, you are doing agentic experience design.

Skills the work needs

  • Interaction design fundamentals. Everything you already know still applies. It needs adapting to a world where the user isn’t the one doing the work.
  • A working sense of how agents behave. You don’t need to know how models are trained. You do need to understand how agents use tools, what they can and can’t remember, and how they can fail differently each time, so you can design for the ways they really break.
  • Thinking about risk. Sorting actions by what they could cost and whether they can be undone is the most important judgment in this job. It has more in common with security work than with visual design.
  • Systems thinking. What you deliver is often a set of rules for permissions, or for when to hand over to a human, rather than a layout.
  • Knowing how AI gets tested. You can’t test something that answers the same question differently each time just by clicking through it. Designers in this role need to be comfortable with running the same task many times, with evals (repeatable tests of AI output), and with failure rates.
  • Writing. Much of what sets the right level of trust is language: how an agent describes what it’s about to do, admits when it isn’t sure, and reports what happened.

Common questions

What does an agentic experience designer do?
An agentic experience designer works out how people hand tasks to AI agents and still stay in charge of the result. That means deciding which actions an agent can take without asking, how it shows what it is about to do, how someone stops or corrects it halfway through, and how it reports what it did afterwards. The job is closer to designing a working relationship than designing a screen.
How is it different from a product designer?
A product designer shapes what people do themselves. An agentic experience designer shapes what happens when an AI model does the task for them. The screen stops being where the work happens and becomes where the work gets approved, watched, and corrected. Most established UX thinking assumes the user does the clicking, and with an agent they often don’t.
Is agentic experience design the same as conversation design?
No, though they overlap. Conversation design is about dialogue: taking turns, prompts, recovering from misunderstandings, and tone. Agentic experience design is about actions taken on someone’s behalf, where the system does things that have real consequences. A conversation designer asks whether the system understood. An agentic experience designer also asks whether the user understood what the system was about to do, and whether they could have stopped it.
What skills does the role require?
Solid interaction design, a working sense of how AI models behave when they use tools (including the way they can fail differently each time), good judgment about risk and what can be undone, and enough systems thinking to design rules for permission and for when to hand over to a human, instead of screens. It also helps to know how AI output gets tested, because you can’t check something that gives different answers to the same input just by clicking through it.
Does the job title exist yet?
Not as a settled title. The work described on this page is real, but there is no agreed name for it yet, so look for the responsibilities rather than the title.

Where to go next

Last reviewed . The title and what it covers are still settling, and we update this page as the role takes shape.

See the path on your own site

The Agent Visualizer follows an AI agent through one of your pages and shows where it arrives, where it reads, where it gets stuck, and everything past that point it never reaches.

Further reading from Field Notes

Posts with more on this topic, each with its date.

All Field Notes

Stay Updated

A monthly email about AI search, AI crawler policy (the rules for the AI bots that visit websites), and new standards for agents. It includes every Field Note published that month, and you can unsubscribe at any time.

We store your email address only to send you this newsletter. See our privacy policy.