AXD vs UX: What Changes When Agents Act for People?

The most critical distinction between Agentic Experience Design (AXD) and User Experience (UX) lies in the assumed state of the human. UX design is fundamentally about the present user. It obsesses over the moments of direct interaction: the clarity of an interface, the intuitiveness of a workflow, the feel of a button click. The user is there, attentive, and engaged with the system, and the goal of UX is to make that interaction as seamless, efficient, and enjoyable as possible.

AXD, in stark contrast, is designed for the absent user. It governs the periods where a human has delegated a task or goal to an autonomous agent and is no longer directly supervising its execution. The core design work is not about the interface the user is touching, but about the trust architecture that allows them to confidently walk away. AXD is about ensuring the agent acts in the user's best interest while they are gone, handling ambiguity, and achieving the desired outcome without requiring constant human oversight.

This shift from presence to absence changes everything. UX optimizes for attention; AXD optimizes for trust. UX designs for interaction; AXD designs for delegation. One is about making the tool easy to use; the other is about making the tool trustworthy to use on its own.

A primary currency of UX design is user attention. The goal is to capture and guide it effectively. This involves creating clear visual hierarchies, minimizing cognitive load, and ensuring that the user can find what they need and complete their task with minimal friction. A successful UX design results in a user feeling that their time and attention were well-spent and respected.

AXD's primary currency is user trust. Since the user is absent, the design focus shifts from managing attention to building and maintaining trust. This is not a soft, emotional concept but a structural component of the system. It involves designing explicit delegation scopes, clear consent horizons, robust recovery mechanisms, and transparent observability. A successful AXD design results in a user feeling confident and secure in the agent's ability to perform on their behalf.

UX design is often focused on enabling users to generate specific outputs. For example, a user wants to create a document, send an email, or purchase a product. The UX is successful if it helps the user produce that output efficiently and with a high degree of quality control over the final product.

AXD, however, is oriented around achieving broader outcomes. The user delegates a goal, not just a series of tasks. For instance, instead of booking a flight (an output), the user delegates the outcome of 'getting to the conference in New York by Tuesday morning as cheaply as possible.' This requires the agent to perform complex decision-making, handle exceptions (like flight cancellations), and chain together multiple actions to achieve the high-level outcome. The designer's job is to structure the delegation in a way that gives the agent sufficient autonomy to achieve the outcome while staying within the user's intended constraints.

The UX designer is an architect of interactive spaces. They use tools like wireframes, prototypes, and user testing to refine the interface and flow of a product. They are deeply involved in the specifics of the user interface and the moment-to-moment experience of the user. Their empathy is directed at the user in the act of using the product.

The AXD designer is an architect of authority and trust. They use tools like delegation frameworks, trust calibration models, and consent architecture to define the relationship between the human and the agent. They are less concerned with the button and more concerned with the contract it represents. Their empathy is directed at the user when they are not present, anticipating their anxieties and ensuring their interests are protected.

You need UX design when a human is actively using a tool to accomplish a task. This applies to everything from websites and mobile apps to complex software and physical devices. If a human is present and interacting, UX is the critical discipline.

You need AXD when you are building a system that can act on a user's behalf without their direct, real-time supervision. This is essential for agentic AI, machine customers, and any autonomous system that requires a significant grant of authority from a human. If the user is meant to delegate and disengage, AXD is the critical discipline.

It is a mistake to see AXD and UX as being in competition. They are complementary disciplines that address different phases of the human-computer relationship. Most sophisticated agentic systems will require both. You need excellent UX for the initial setup, the delegation process, and the review of the agent's performance. This is the 'present user' phase where the human interacts directly with the agent's interface.

Once the delegation is made and the user is absent, the principles of AXD take over, governing the agent's autonomous operation. A successful system is one where the UX of the delegation interface is so clear and the underlying AXD is so robust that the user feels completely comfortable turning their attention elsewhere. The ultimate goal is a seamless transition from a state of presence, managed by UX, to a state of absence, secured by AXD.