Your Next User Won't Have Eyes. Are You Ready?
AI agents are quietly becoming users of the products we build. The design equation has changed — it's no longer just about designing for humans, but for humans and agents together.
We built UX around humans.
The assumption was clear: the user is a person. A person with emotions, biases, motivations, habits, frustrations, goals, and social context.
So the foundations of UX were naturally built on:
- —Psychology
- —Sociology
- —Behavioral science
- —Cognitive science
That made sense. Because the design problem was: how do we help humans interact with systems better?
Now that assumption is breaking.
Agents have entered the workflow.
AI agents that can read, decide, and take action on someone's behalf are quietly becoming users of the products we build. They book meetings, send emails, write code, pull data, fill forms, and increasingly hand off tasks to other agents.
That changes the design equation.
The system is no longer being used only by a human. It is now being shaped by a human-agent partnership.
So the question is no longer, How do we design for human use?
It becomes, How do we design for humans and agents?
That means the old foundation is no longer enough.
The New Foundation
Agents are not human users. They do not have emotion, identity, social anxiety, aspiration, or intuition in the human sense.
They operate through:
- —Goals
- —Instructions
- —Context
- —Permissions
- —Memory
- —Constraints
- —Feedback
- —Confidence levels
- —Exception handling
So to design for them, we need a different design lens.
A clear map of what things are, not just how things look. Designers used to arrange information so people could find it with hierarchy, spacing, and typography. Agents don't see any of that. They need the structure underneath. A clear inventory of what exists in your product, what each thing is called, and how the pieces relate to each other. A lot of what used to feel like how pretty does this look? is joined by something much closer to library cataloguing — is this labelled in a way a machine can make sense of?
Clear actions, not clickable buttons. People interact with products by tapping, swiping, and clicking. Agents interact by calling functions. Specific, named actions the product exposes to them. That means the real interface for an agent isn't your screen — it's the menu of things your product lets it do, how those actions are described, and what happens when they go wrong. Designing that menu well is now part of the product experience, even though no human will ever see it.
Rules for trust, not rules of etiquette. When people interact with each other through a product, we use social design — community guidelines, reputation systems, polite defaults. When agents interact with each other (often on behalf of different people), we need something sturdier: clear rules about who's allowed to do what, on whose behalf, and with what limits. What happens when an agent books a flight you didn't approve? What happens when two agents negotiating for two humans reach a deal neither human actually wanted? These are new design problems, and they feel closer to writing the rules of a game than decorating a living room.
Honesty about what the system knows. Agents make things up. Not always, not maliciously, but often enough that it matters. So a new design responsibility appears: helping users see where information came from, how confident the system actually is, and where it might be guessing. This used to be a small detail — a footnote, a "source" link. Now it's central. A product that can't show its work is a product that can't be trusted with anything important.
So What is the New Base?
It's easy to read all this and conclude: "So human UX is dead, long live agent UX."
It isn't, and it won't be.
The new base is not a replacement for psychology and sociology. It is an expansion.
It becomes: Designing for people, intelligent systems, and the relationship between them.
In the real world, agents rarely work alone. A person asks an agent to do something. The agent does part of it. The person checks the result. Another agent might step in. A second person approves. The user isn't a human or an agent. It's the handoff between them.
Which means the old design foundations aren't going anywhere. They still apply on the human side of every handoff. The new foundations apply on the agent side. And the most interesting, least-explored territory is the joint between them. And I think that joint is where the next decade of design work actually happens.
What Designers Must Learn Next
If design is going to stay relevant, we need to expand our foundation. Not abandon human-centred design. Expand it. Because we are no longer designing only for human users. We are designing for a new working unit — the human-agent team.
So we need to redefine Human Centred Design (HCD) to Human & Agent Centred Design (HACD).
A few shifts are already underway and worth watching.
The title "UX designer" is going to stretch. Some designers will specialise further in human experience. Others will move toward designing for agents — a space some are starting to call AX, or Agent Experience. Most valuable roles in the middle will require both.
I actually saw a design job post on LinkedIn for Agent Experience Designer.
Documentation and help articles — long treated as afterthoughts — suddenly become primary. For human users, docs support the product. For agents, the docs are the product. They're the main way an agent figures out what your product does.
Research methods need to change, too. Watching people use your product tells you something real. But it won't tell you whether an AI can reliably use your product. That takes a different kind of testing — stress-testing the system with agents themselves, looking for where they misunderstand or break things.
And design systems. The rulebooks that keep products consistent are going to quietly absorb things that used to belong to engineering: how data is named, how actions are labelled, how errors are reported. Because if your buttons are beautifully consistent but the actions underneath them aren't, half your users are having a broken experience.
The users changed. The foundations have to follow. And the designers who move first won't just adapt to the shift. They'll define it.
Share this article
Related Articles
Design Leadership in 2026: The New Role of Design Leaders
The role of a Design Leader has changed dramatically. Here are six things every design leader needs to lean into — from AI as a design assistant to empathy as a core leadership skill.
The Rush to Lead: Why True Leadership Takes Time, Patience and Scars
Many early professionals want to become a lead within one or two years. Here's a straight-from-the-heart look at why that rush happens, what leadership really demands, and how to build the habits that make you a natural leader.