The new elements of UX
In 2000, Jesse James Garrett published The Elements of User Experience and gave the design profession something it badly needed: a shared language for how design participates in the development of a product. His five planes — Strategy, Scope, Structure, Skeleton, and Surface — clarified how design decisions made at each layer shape the ones above it, and gave design leaders a framework for arguing that design's involvement needed to begin at the strategy level, not the surface.
It was the right model for its moment. That moment has passed.
New planes of Generative UX
Generative AI doesn't just change what the surface looks like. It changes the nature of every layer beneath it. The interface is no longer fixed. The information architecture is no longer static. The user is no longer a defined persona with a predictable workflow. The system's capabilities are no longer stable between releases. And the data that powers all of it is no longer a warehouse to be queried — it is a living substrate that shapes what the system can do, who can do it, and what it means.
Garrett's five planes were built for a world where design defined the experience in advance. We are now designing systems that define themselves in real time. That requires a new model — not a replacement for Garrett's thinking, but an evolution of it, built for the complexity of AI-mediated enterprise systems and the design leaders responsible for shaping them.
These are the seven planes of Generative UX.
1. Multi-Modal Surface: How the system shows up
For fifty years, the surface of a digital product has meant one thing: a screen. Pixels arranged into interfaces, navigated with a mouse or a finger. Garrett's Surface plane assumed this so completely that it didn't need to say it.
That assumption is gone.
The surface of a generative system can be visual, auditory, haptic, or spatial — and increasingly, it is all of these at once. A field worker interacting with an enterprise system through voice commands while their hands are occupied. A manufacturing environment where gestures replace clicks. A trusted set of orchestrated agents performing tasks with no interface at all, surfacing only their outcomes and the audit trail that confirms they operated within sanctioned boundaries.
For design leaders, this means the Surface plane is no longer about how the product looks. It is about how the system shows up — in what form, through what modality, in what physical and organizational context. The design question is no longer "what does this screen look like?" It is "how does this system manifest itself to this person, in this moment, in this environment?" And increasingly, the answer will be different every time.
2. Conversational Context: How meaning is established
Garrett's original model had no plane for language. In consumer web design, that was a reasonable omission. In enterprise software, it was always a gap — domain-specific terminology, instructional copy, error messages, and onboarding language are critical design decisions that shape whether a system feels intelligent or alienating, trustworthy or opaque.
Generative AI makes this gap impossible to ignore.
In a generative system, language is not static content to be written once and reviewed annually. It is a dynamic layer that the system produces in real time, calibrated to who the user is, what they are trying to do, what they have done before, and what the system anticipates they will need next. The terminology that helps a new user understand a workflow becomes clutter for an expert. The instruction that is appropriate for a standard transaction may be insufficient or misleading in an edge case with compliance implications.
This plane is about the design of understanding itself: how the system tracks intent, manages context across a conversation, and produces language that is not just accurate but situationally appropriate. For design leaders, it means taking ownership of how the system communicates — not just what it says, but when, to whom, and in response to what. That is a design problem. It should not be left to engineers or content strategists working without design leadership.
3. User Aspiration & Evolving Models: Who the user is becoming
Garrett placed user segments in the Strategy plane, which was appropriate for consumer products where personas are relatively stable. In enterprise software, the concept of roles has always done heavier lifting — defining not just who the user is, but what they can see, what they can do, and what they are accountable for.
Generative AI stretches this further than role-based access control was designed to handle.
The users of an AI-mediated enterprise system are not static. They are developing. A system that understands where a user is in their professional development — what they know, what they are learning, what they are ready for — can do something no previous enterprise system could: grow with them. Adaptive learning environments, proactive enablement, career-aware nudging, role progression based on demonstrated capability rather than job title. The system isn't just responding to who the user is today. It is participating in who they are becoming.
For design leaders, this plane reframes the user research question. It is no longer enough to understand the user's current needs and workflows. The design question becomes: what does this person need to be effective today, and what does the system need to understand about them to support where they are going? That is a richer, more ambitious brief — and it is one that design is uniquely equipped to answer.
4. Generative Hierarchies: How structure adapts in real time
Information architecture has always been one of design's most powerful and least visible contributions. The decisions made at this layer — how content is organized, how navigation is structured, how relationships between pieces of information are expressed — shape the user's entire experience of a system's intelligence and coherence.
In a static system, these decisions are made once and revised periodically. In a generative system, they are made continuously, in real time, by the system itself.
AI-powered systems generate and prioritize informational hierarchies based on context: the user's intent, the data being accessed, the organizational policies governing what can be shown to whom, and the classification of the output being generated. A workflow that is appropriate for one user in one context may be structured entirely differently for another user in a different role accessing the same underlying data. The architecture floats — it responds to what is generated rather than constraining it in advance.
This has significant implications for governance. When a new dataset generates new capabilities — new IP, new predictions, new classifications that didn't exist when the system's roles were defined — access to that information cannot be governed by a role that was created before the capability existed. The architecture must be designed to adapt.
For design leaders, this plane is where the argument for design's involvement in data governance begins. The decisions made about how information is structured, classified, and surfaced are design decisions. They shape the user's experience of the system's intelligence as directly as any visual choice. Design needs to be in that conversation.
5. Inferred Intention & Expectations: What the user means
The gap between what a user says and what a user means has always been a design problem. In a static system, it was addressed through clear labeling, sensible defaults, and error prevention. In a generative system, it becomes the central design challenge.
This plane is the predictive core of generative UX. The system infers what the user is trying to achieve by blending behavioral signals, prompt patterns, past actions, and role-based heuristics. The experience becomes anticipatory: accelerating likely workflows, generating starting points before the user asks for them, alerting to blockers before they become problems. When the inference is right, the system feels intelligent and trustworthy. When it is wrong, it needs to handle the correction gracefully — without losing the user's confidence.
This is also where traditional role-based access control breaks down most visibly. RBAC was designed to govern static screens, fixed functionality, and predefined data sets. In a generative environment where all of these are produced in real time, RBAC's assumptions no longer hold. Access governance needs to shift toward models based on the intent of the request, the attributes of the data being accessed, and the context of the interaction — including the physical setting, the other people present, and the risk level of what is being generated.
For design leaders, the implication is significant: the design of access and governance is now inseparable from the design of the experience. These are not separate workstreams. They are the same conversation, and design needs to be leading it.
6. Emergent Capabilities: What the system is becoming
In every product development model before this one, the system's capabilities were defined in advance. Features were scoped, designed, built, tested, and shipped. The surface changed with each release. The underlying capability set was stable between them.
That is no longer true.
A generative system's capabilities evolve continuously: through model updates, plugin ecosystems, retrieval-augmented generation drawing on updated private data, and the refinement of intent recognition accelerated by the behavioral patterns of hundreds of enterprise users repeating the same workflows every day. The system that exists on Monday is meaningfully different from the system that existed on Friday — not because anyone shipped a release, but because the system has learned.
For design leaders, this is one of the most disorienting shifts in the new model. Predefined workflows are becoming legacy artifacts. Emergent workflows — ones that adapt to the collective behavior of teams as well as to individual prompts and requests — are becoming the norm. UX design in this plane is less about defining the workflow and more about curating the prompts and parameters that shape what the system generates in response to it.
This plane also carries the heaviest governance burden. As capabilities emerge, the system must be able to explain what it can do, expose how it is doing it, and safely extend what it offers while managing change, risk, and transparency. In enterprise contexts, innovation and control are not opposites — they are design constraints that must be held in balance simultaneously. That balance is a design leadership responsibility, not just an engineering one.
7. Data Intelligence: What makes the system trustworthy
Every plane above this one rests on this one. The quality of the surface, the coherence of the conversation, the accuracy of the inference, the integrity of the governance — all of it depends on how data is collected, structured, governed, and used. This is not a technology problem that design can safely leave to engineers. It is a design problem that happens to require engineering to solve.
In a generative enterprise system, data is no longer optimized for human consumption or static workflows. It needs to be structured and governed for agents operating in adaptive, real-time contexts. That means moving from warehousing models to live, contextually available data with semantic layering. It means augmenting human-readable metadata with machine-optimized representations — embeddings, entity relationships, intent signals — that allow the system to retrieve the right information, fast, with confidence scoring and traceability. It means replacing monolithic pipelines with modular, composable integrations that can respond to the specificity of a prompt rather than the generality of a schema.
Most importantly, it means dismantling the data silos that currently permeate enterprise organizations. The new version of Conway's Law is this: you will be able to read a company's org chart in the fractures of their data strategy, and those fractures will show up directly in the coherence — or incoherence — of their generative user experience. Data fragmentation is a UX problem. Design leaders who understand this will have influence far beyond the interface.
This is the plane where trust is either built or forfeited. A system whose data is well-governed, explainable, and traceable is one users can rely on. A system whose data is opaque, siloed, and ungoverned is one that will eventually betray the people using it — regardless of how well-designed the surface appears to be.
The floating foundation
Garrett's five planes gave design a model for participating in the development of products from strategy to surface. That contribution was enormous and its influence on the profession cannot be overstated.
But those planes were built for a world where the experience was defined in advance and delivered as designed. We are now building systems that define themselves in response to the people using them — systems that learn, adapt, infer, and generate in real time. The experience is no longer a fixed artifact. It is an emergent property of the system's intelligence, the user's intent, and the data that connects them.
Like a building on a floating foundation, generative user experiences achieve stability not by resisting change but by responding to it. They do not rest on a static hierarchy of layers. They rest on the dynamic equilibrium of evolving data, evolving users, and evolving capabilities — adjusting continuously to preserve coherence, relevance, and trust.
Design leaders who understand this will not be caught defending the primacy of the graphical interface while the ground shifts beneath it. They will be the ones defining what comes next.
Appendix: Garrett's Five Planes
Jesse James Garrett's The Elements of User Experience defined five planes of UX design, each building on the one below it. While the model requires updating for the age of generative AI, its foundational contribution to design thinking remains significant.
Surface: How the product looks and feels. This plane focuses on visual elements — typography, color, imagery — ensuring the product is aesthetically coherent and easy to understand.
Skeleton: How users interact with and navigate the product. This plane focuses on interface elements such as buttons, menus, and navigation systems, optimizing usability and streamlining user flow.
Structure: How features and content are organized. This plane defines the underlying framework of the product, arranging information and functionality into a coherent system to support intuitive interaction.
Scope: What the product includes and excludes. This plane translates strategic objectives into specific features and content, defining the product's functional boundaries and capabilities.
Strategy: Why the product exists. This foundational plane defines the product's goals by aligning business objectives with user needs. Its purpose is to establish a clear vision for success.
