For most of my career, the user was clear. A person, on a screen, is trying to get something done. That assumption is starting to break. We are entering a phase where designers are no longer optimizing interfaces only for humans, but for AI agents that act on behalf of humans. And that changes the foundation of UX.
In 2026, AI is no longer just assisting. It is acting. Agentic systems are moving from passive tools to autonomous actors that plan, execute, and iterate on tasks without constant human input. Instead of a user clicking through a workflow, an agent interprets a goal and moves through the system on their behalf. The question shifts from “How does a user navigate this?” to “How does an agent complete this?” That is a fundamentally different design problem.
Traditional UX was built around human behavior. Visual hierarchy, affordances, microinteractions, and progressive disclosure were designed to support how people perceive and decide. AI agents do not need any of that. They do not rely on visual cues or interaction patterns. They rely on structured information, predictable systems, and clear, machine-readable actions. They operate closer to APIs than interfaces. That creates tension, because we are now designing systems that must work for both humans navigating and agents operating, and those needs are not always aligned.
We are already seeing early signs of this shift. Interfaces are becoming more structured, more predictable, and more consistent. Not for aesthetics, but for execution. Actions are labeled more clearly, flows are standardized, and ambiguity is being reduced. At the same time, AI is beginning to generate interfaces dynamically based on user goals, which means the UI is no longer fixed. The interface becomes temporary. The system becomes the product.
This is where UX starts to move. Designers are no longer just shaping screens, they are shaping systems that agents can operate within. That includes defining actions clearly, structuring data for interpretation, designing predictable workflows, and building in error handling and recovery. Because agents will fail, and when they do, the experience has to recover in a way that still makes sense to the human overseeing it.
There are real risks in this transition. If systems are optimized too heavily for agents, human clarity suffers and the experience can feel mechanical or opaque. If systems are optimized only for humans, agents struggle to operate and automation breaks down. The balance between those two is where the work is, and most teams are not thinking about it yet.
This is part of a larger shift. We are moving from interaction to delegation, from interfaces to outcomes, and from user experience to something closer to system experience. Users are no longer the only actors. Agents are increasingly doing the work. The human role becomes setting intent, supervising outcomes, and making decisions when needed.
The bottom line is simple. Designers are no longer just designing for people. They are designing for systems that act on behalf of people. That requires a different mindset. Not how the user clicks through a flow, but how the system understands, decides, and acts. Because in a world where agents are doing the work, the best experience is not the one that is easiest to use. It is the one that is easiest to operate.