NN/g: Experience Design – dropping the 'user' from user experience
Don Norman coined the phrase “user experience” in the 1990s to describe something broader than usability: not just the interface, but the whole of a person’s encounter with a product or service. AI agents are now straining that framing. When a personal AI acts on a person’s behalf — booking a table, filing a form, managing a schedule — the agent is the one making requests, while the person waits for a report. Raluca Budiu’s October 2026 article for Nielsen Norman Group proposes “experience design” as the name for the larger discipline that this shift requires.
The premise is not that UX is obsolete. The people-centered core of UX now needs to hold across a wider set of actors and interactions, and the existing terminology does not cover that scope cleanly.
Three areas of scope
Budiu outlines three distinct situations that experience design must now address:
- Humans using products directly — the classic UX case.
- Agents acting on behalf of humans — the new case, where the agent is the proximate user and the human is one step removed.
- Humans and agents working together inside the same product — hybrid flows where control passes back and forth.
The third situation generates what the author calls handoff problems: a task passes from agent to human (or the reverse), and gaps in continuity, authorization, or recoverability open up. The design problems this creates — what should the agent remember, when should it ask, when should it act, how can a person inspect or undo what it did — are not covered by traditional user flows or task analysis.
Methods and measurement
AI speeds up many parts of the traditional UX toolkit, and it enables approaches like synthetic respondents for initial testing. But agents are also new objects of study. How users experience an agent-mediated service may require adapted evaluation methods, and the author argues that qualitative research stays important — agents often perform well on task completion metrics while users feel uncertain or uninformed about what happened.
The fundamental success criterion does not change: did the person reach their goal? Designing for agents does not mean optimizing for agent success alone.
Who this is for
The article is useful for UX practitioners and researchers starting to work on agent-accessible products or AI assistants, for product teams building services that agents will call on behalf of users, and for design leaders updating their team’s scope and methods as AI takes on more of the interaction layer.