English | EN 
For a long time, UX work in digital products has revolved around a seemingly obvious question: Is this easy to use? That question has shaped our methods, processes, and competencies. We have measured efficiency, reduced friction, refined flows, and strived for predictability. This made sense, because the systems we built were by and large predictable. UX was about making the flow simple, understandable, consistent, and efficient for the user.
AI is, as you know, not new. For years it has been used to generate code, analyze data, write tests, or streamline design work. In these cases, AI has been a powerful tool in the creation of a solution, but not as the solution the user ultimately interacts with. The deliverable has still been a traditional digital product: a system, an interface, a website, a flow. As a result, the UX question could remain largely untouched, even though the way we achieved delivery evolved.
What truly changes the game is when AI is no longer just used within the delivery process, but when AI is the deliverable. When it is the AI that the user meets and collaborates with. When the solution is no longer a fixed flow encapsulated in code, but a system that interprets, generalizes, and responds with a degree of uncertainty. This is where the classic UX question enters unfamiliar territory. It shifts from “Is this easy to use?” to “How should the AI interact with the user?”
This new question is not primarily about UI, but relationships. It is not about individual flows, but about interaction over time. An AI‑based deliverable does not behave the same way every time. The same question can yield different answers depending on context, phrasing, or previous interactions. Where traditional systems were built on a clear contract between human and machine, with transparent cause and effect, AI introduces a more open and probabilistic form of interaction. UX must treat this as the normal state, not as an exception.
In this context, it also becomes clear that AI deliverables are never finished in the classical sense. They continue to evolve, adjust, and improve after they are put into use. The boundary between exploration and delivery becomes blurred. Exploration ceases to be a clearly defined phase before delivery and instead becomes a continuous condition. Delivery is no longer about shipping a stable and final product, but about ensuring continuous learning within clearly designed constraints.
This has direct consequences for the UX role when AI is the deliverable. It is no longer sufficient to ensure that an interface is understandable or that the user quickly reaches the end of a flow. UX must also take responsibility for how the AI behaves during its encounters with users. When should the AI take initiative, and when should it hold back? How does it communicate uncertainty without undermining trust? When should it ask questions rather than provide answers? These concerns sit at the intersection of design, ethics, and risk management, and are fundamental UX questions.
Many teams have spent recent years exploring AI through rapid experiments and quick proof‑of‑concepts before development and delivery. When AI becomes the deliverable, however, the focus shifts from code to behavior. What is being explored is not only whether something works technically, but how the AI is experienced, interpreted, and how it influences user decisions. Prompts, examples, specifications, and constraints become design material in the same way sketches and prototypes have been in traditional UX. The critical difference is that what is being explored is not a prototype, but what users encounter in real life.
When AI is the deliverable, UX becomes the discipline that ensures that this advanced, learning system provides a human‑adapted output. UX does not eliminate uncertainty but makes it understandable. UX does not attempt to make AI perfect but makes it clear to users how it should be engaged with. The focus shifts away from optimizing individual features toward establishing a sustainable interaction between human and machine, with a clear outcome in mind.
For those of us working with innovation, development, and advisory services, this represents a clear shift in perspective. AI will continue to be a powerful tool during the development phase of traditional solutions. But when AI becomes the deliverable itself, the conditions fundamentally change. At that point, it is no longer enough to ask whether a solution is easy to use. The decisive question becomes how the AI interacts with the user over time, in uncertainty. It is in this shift that UX moves from a supporting function to a foundational capability in creating AI‑driven deliverables.