By Germán León
I stopped writing about AI and UX for a while. Not because I lost interest, but because it became almost impossible to write about it with any clarity. Every time I thought I understood where things were going, the landscape shifted again.
One week, it was chatbots. Then agents. Then multimodal systems. Then the copilots. Then image generation, video, voice, and memory. Now we have canvases, projects, artifacts, workflows, assistants, all overlapping, all partially defined, all competing to describe the same emerging category.
For a time, writing about AI felt like reviewing a building while the architects were still moving the walls.
But something has changed. Not the pace of innovation, that is still accelerating. What has changed is that the underlying problem is finally becoming visible.
For the last two years, the dominant narrative in AI has been about intelligence. Which model is smarter, which one reasons better, which one codes better, which one generates higher quality outputs? That race has been real, and it has mattered.
But it has also been incomplete.
Because for most people today, intelligence is no longer the main bottleneck. AI is already good enough at many tasks. The real friction appears somewhere else, somewhere less glamorous and far more decisive.
The bottleneck is the experience.
Most AI products today are powerful, but they are also cognitively demanding. They require users to navigate an invisible layer of complexity before they can extract value. Users are asked to understand models, modes, tools, files, permissions, memory systems, workflows, connectors, and product logic that is rarely explicit.
This is not a small detail. It is a fundamental design failure.
AI was supposed to reduce complexity. Instead, many tools have introduced a new kind of complexity, one where the user must not only know what they want, but also understand how the machine expects to be used. The interface becomes a negotiation between human intention and system constraints.
That is not good design. That is raw capability leaking into the product surface.
You can see this tension clearly when looking at the major players.
Google, for example, has everything it needs to dominate this space. It has the infrastructure, the research, the distribution, the browser, the operating system, the productivity suite, and billions of users. And yet, from a product experience perspective, its AI ecosystem feels fragmented.
There is Gemini, NotebookLM, AI Studio, features embedded in Workspace, and a range of experimental tools. Each one is powerful in isolation, but they do not yet form a coherent experience. Users do not feel like they are interacting with a unified system. They feel like they are navigating a collection of disconnected entry points.
This has been a recurring pattern for Google. Exceptional technology, inconsistent product orchestration. In the context of AI, that gap becomes more visible because users need more than functionality. They need orientation, clarity, and trust. They need to understand where they are inside the system, what the system can do, and how different parts relate to each other.
Good UX is not a collection of features. It is orchestration.
OpenAI sits in a different position. It deserves enormous credit for making AI accessible at a global scale. ChatGPT fundamentally changed how people think about software. It turned interaction into conversation, and it lowered the barrier to entry in a way that had not been done before.
But success introduced a new challenge.
The product has become extremely powerful, but also increasingly complex. There are multiple models, tools, modes, custom GPTs, projects, canvases, memory systems, file handling, voice interaction, and connectors. For advanced users, this is exciting. For most users, it creates friction.
The issue is not capability. It is navigation.
The product's surface is expanding faster than the user’s mental model can keep up. And when that happens, something subtle shifts. Users stop focusing on what they want to create and start focusing on how to operate the tool itself. They begin searching for features rather than ideas.
That is a dangerous moment in product design. It signals that the system is no longer intuitive.
Across the broader market, especially in fast-moving ecosystems like China, the focus remains heavily on output. Faster generation, better video, more impressive demos. The progress is undeniable, but much of it remains oriented toward showcasing capabilities rather than defining a repeatable way of working.
A product can generate something impressive and still fail as a daily tool. Output is not the same as usefulness. The companies that will win in the long term are not only those that produce the best results, but those that embed themselves into real workflows.
This is where Claude begins to feel different.
Claude is not perfect, but it is the first AI product that clearly reflects an important shift in thinking. It treats interaction as the core of the product, not intelligence alone.
It feels calmer. More legible. Less performative. It does not constantly try to demonstrate its capabilities. Instead, it behaves more like a tool designed to help you move forward. The emphasis is not on impressing the user, but on supporting the process.
That distinction is subtle, but it changes the entire experience.
With Claude Design, this becomes even more evident. The value is not simply in generating visuals. That capability is becoming commoditized. What matters is the workflow it enables. The user moves from intention to artifact, from artifact to iteration, and from iteration to something closer to a system.
It is not about producing a single output. It is about enabling thinking.
This shift has significant implications for design as a discipline.
For decades, design workflows started with a blank canvas. Designers opened tools like Figma or Photoshop and began constructing artifacts from scratch. Now, the starting point can be language. Language becomes a way to define direction, constraints, tone, and intent before any visual is produced.
As a result, the designer's role is moving upstream.
Less time is spent generating the first version. More time is spent defining what “good” looks like. Designers are increasingly responsible for setting direction, evaluating outputs, maintaining coherence, and translating vague ideas into structured systems.
This is not a reduction in the importance of design. It is an expansion.
When anyone can generate an interface, the world becomes saturated with interface-shaped objects. Many of them will look polished. Many of them will be meaningless. AI can generate the appearance of design, but it cannot guarantee relevance, usability, or clarity.
That responsibility remains human.
This is why UX is not becoming less important. It is becoming more critical. The cost of poor judgment increases when production becomes easier.
The same applies to design systems. In an AI-native workflow, design systems are no longer just documentation for human teams. They become inputs for machines. They define the structure within which AI operates. A clear system produces coherent outputs. A weak system amplifies inconsistency.
In that sense, design systems become part of the organization’s infrastructure. They make the product legible not only to people, but also to the machine.
At the same time, the acceleration of generation introduces a new risk. As it becomes easier to produce content, it becomes easier to scale mediocrity. We are already seeing an increase in outputs that look finished but lack depth, strategy, or connection to real user needs.
The internet is filling with convincing, well-executed, but ultimately empty artifacts.
This is where education and practice need to evolve. The focus cannot be on tools alone. It must shift toward judgment, problem framing, evaluation, and systems thinking. The future designer is not defined by their ability to generate, but by their ability to decide.
What should exist.
What should not.
What works.
What does not.
For the last two years, AI has been defined by the model race. That phase is not over, but it is no longer the only game.
We are entering the product phase.
The central question is changing.
It is no longer just about which model is smartest.
It is about which product allows people to think, create, and act with the least friction and the most trust.
That is a UX question.
And right now, most of the industry is still answering it poorly.
Claude is not the final answer. But it is the first clear signal of a different direction. A direction where AI is not just powerful, but usable. Not just impressive, but integrated into real work.
In a space where everything is moving quickly, direction matters more than perfection.
And for the first time in a while, that direction feels visible.
The future of AI will not be won by intelligence alone.
It will be won by experience.