There is a date in design history that nobody marks: the date the screen stopped being the primary site of value creation in digital systems. We cannot name it precisely because it did not happen on a particular day. It happened gradually, then suddenly, in the way that most structural shifts do. But it has happened. And the discipline of UX - built entirely around the screen, the tap, the scroll, the journey - is now designing for an experience that is no longer the dominant one.
This essay is not a eulogy for UX. UX will remain necessary for the foreseeable future. Most digital products still have screens. Most digital experiences still require a human in the loop, navigating a designed interface. The skills that UX has accumulated over thirty years - cognitive psychology, information architecture, interaction design, usability - remain genuinely valuable. As the AXD Manifesto argues, this is not a replacement but a parallel discipline.
But they are no longer sufficient. And the gap between what UX can do and what the moment requires is widening faster than the discipline is moving to close it.
The shift is not from one interface paradigm to another. It is from interface-centric design to something more fundamental: the design of relationships between humans and systems that act.
Three Transitions, Three Disciplines
To understand where we are, it helps to trace how we got here. Digital design has undergone three structural transitions, each of which required not just new tools but new conceptual foundations.
The first transition was from command-line to graphical interface. The desktop metaphor - folders, files, the desktop itself - was a profound act of translation. It took the abstract logic of the computer and made it legible to human intuition. The design discipline that emerged from this transition was Human-Computer Interaction: precise, laboratory-founded, concerned with efficiency and error rates. It was a discipline built for the computer as a tool.
The second transition was from desktop to web, and then to mobile. This required a new discipline because the medium was fundamentally different: distributed, connected, variable in display, navigated rather than operated. UX was the answer. It brought ethnographic research, journey mapping, and customer empathy into the design process. It was concerned not with efficiency but with experience - the arc of feeling a person traversed while using a digital product. UX was a discipline built for the screen.
In the first two transitions, the computer waited. It responded when asked. The human was always in the loop - executing the flow, navigating the journey, completing the transaction. The designer's job was to make that loop as frictionless, legible, and satisfying as possible.
In the agentic transition, the system acts. It initiates. It operates in the world while the human is absent. It makes decisions, executes transactions, coordinates with other systems, and returns not an interface state but a result. The loop - the interaction loop that UX was built to optimise - no longer exists as the primary structure of the experience.
What Breaks
When you try to apply UX methods to agentic systems, you discover immediately what breaks. Journey mapping breaks - because there is no journey, only an outcome. Interaction design breaks - because there are few interactions, only delegations and interruptions. Usability testing breaks - because you cannot observe the primary use state, which is absence. User flows break - because the agent chooses the path, not the user.
The deep, empathic understanding of human needs, fears, and mental models remains essential. The skill of listening to what people actually want, not what they say they want, does not become less important when the interface disappears - it becomes more important.
What survives? Research. The deep, empathic understanding of human needs, fears, and mental models remains essential. Because in an agentic system, the human's expressed intent is the only design input the designer has to work with. If that intent is misunderstood, the agent will act on the wrong goal. There is no interface for the human to correct course in real time.
Systems thinking survives. The ability to reason about complex, interconnected systems - the skill that service design brought to UX - becomes central rather than peripheral in agentic design. Multi-agent systems are systems of systems. They require a designer who can think at multiple altitudes simultaneously.
Ethics survives - or rather, ethics becomes unavoidable in a way it was previously possible to defer. In agentic design, an agent acting on incorrect, biased, or unsafe instructions does immediate, concrete damage. The ethical architecture of an agentic system is not an afterthought - it is a design prerequisite.
The Trust Problem
At the centre of all of this is a problem that UX has never had to solve at scale: the trust problem.
Trust, in UX, was a user experience quality - something designers aspired to create through consistency, clarity, and brand. It was important, but it was not structural. A poor trust experience meant a user would leave the app or choose a competitor. The consequences were commercial, not operational.
In agentic systems, trust is structural. It is the mechanism by which autonomous action is authorised. An agent that a human trusts completely will be given wide scope to act - to transact, to commit, to obligate. An agent that loses trust will be constrained, monitored, or disabled. The level of trust a human has in an agent determines the operational envelope within which the agent can function. This concept is explored in depth in Trust Architecture, where we examine the four layers of trust in agentic AI systems.
Trust in agentic systems is not a feeling. It is an architecture. And it can be designed well or badly, with consequences that are immediate, measurable, and in regulated industries, legally significant.
This is where the intersection with financial services becomes so revealing. Banks are institutions of trust. They have spent decades - centuries - building the regulatory, reputational, and experiential infrastructure that allows people to delegate financial authority to them. The design problems are not new. They are ancient. But agentic AI makes them urgent at scale and at speed that the existing infrastructure was not designed to handle.
The 18-Month Window
There is a window. It is approximately 18 to 24 months in duration. It opens now and it closes around Q2 2027. During this window, the architecture of human-agent experience in high-trust domains - banking, healthcare, legal, government - will be set. The patterns that are established in this window will calcify into defaults. The vocabulary that is developed in this window will become the vocabulary of the discipline.
Currently, the answer is predominantly engineers and product managers working at AI companies. They are building agentic systems with extraordinary technical sophistication and very little design theory to guide the human experience side of the equation. The results are what you would expect: technically impressive, humanly unreliable.
Design has an 18-month window to enter this conversation at the foundational level - not as an aesthetic afterthought applied to AI products, but as a co-equal discipline that shapes what agentic systems are designed to do, how they earn and maintain trust, and how they fail gracefully when they fail.
What the Discipline Looks Like
Agentic Experience Design - AXD - is the name for this discipline. It is not a rebrand of UX. It is not a specialisation of UX. It is a parallel discipline that will collaborate with UX in the same way that service design collaborates with UX: sometimes overlapping, sometimes in tension, always serving a set of design challenges that the other cannot fully address alone.
AXD is concerned with delegation architecture: how humans grant authority to agents, what scope that authority covers, and how it can be modified or revoked - a topic explored fully in Delegation Design. It is concerned with trust calibration: the ongoing negotiation between human confidence and agent reliability that determines the operational envelope of autonomous action.
It is concerned with interrupt design - the moments when agents must surface decisions back to humans - which are arguably the most high-stakes interaction design challenges in the field. An interrupt that comes too often trains humans to dismiss it; an interrupt that comes too rarely allows consequential errors to accumulate before a human can intervene.
And it is concerned with what I call the Invisible Layer: the stratum of agentic capability that creates value without any visible interface activity at all. Banking services that are negotiated, optimised, and executed on your behalf while you sleep - by machine customers acting on your behalf. Insurance that adjusts its parameters to your real-time life context without requiring you to fill in a form. These are not science fiction. They are products being built today. The AXD Practice section provides frameworks for designing these systems.
A Discipline That Earns Its Authority
New disciplines do not announce themselves - they are built, term by term, framework by framework, case study by case study, until they have accumulated enough rigour to be taught, enough vocabulary to be cited, and enough practitioners to be a community.
The goal is not to be the loudest voice in the conversation about AI and design. The goal is to be the most precise, the most grounded, and the most useful. To build the frameworks that practitioners actually reach for when they are designing a trust architecture or a consent system or an interrupt pattern.
After the interface, there is not nothing. There is design that is harder, more consequential, and more necessary than anything we have done before. That is the territory AXD is here to claim.
The screen served us well. It will continue to serve us. But the primary challenge of digital design has moved. It has moved to the relationship between humans and systems that act - to the invisible layer, the trusted agent, the delegated authority.
That territory has a name now. It has a vocabulary, a set of founding principles, and a community gathering around it. It is called Agentic Experience Design.
The discipline starts here.