(ML)
DESIGN LEAD
Every few months, someone posts a thread showing how they built a fully realized app over a weekend—polished screens, coherent flows, working interactions—using little more than AI tools and a stack of prompts. The reaction is immediate and familiar: product designers are next.
There is a reason those demonstrations are persuasive. They are fast, visually impressive, and often more complete than most observers would have thought possible even a year ago. But they share a consistent omission. They show the screen. They do not show the conversation that had to happen before the screen existed.
That conversation is the one where someone determines which problem is actually worth solving, for which users, under what constraints, and in service of what business outcome. That work does not disappear because a tool can generate an interface. It is not, at its root, a prompt-writing problem. It is a judgment problem. And judgment has always been the real center of product design.
A great deal of the current discourse conflates the visible artifact with the discipline itself. That is an understandable error. Screens are easy to see and easy to evaluate. Decisions are not. But any serious assessment of what AI means for product designers has to begin by separating the two — by asking what product design actually is, rather than what it most visibly produces.
Most anxiety about AI replacing designers centers on the most visible layer of the work: the interface. A screen is easy to generate, critique, and share online. But product design has never been reducible to interface production, and the assumption that it could be has always been a misreading of the discipline.
The real work happens upstream, in the messy and often invisible space between a business problem and a design artifact. It is where designers reconcile competing realities: what users need, what the business is trying to achieve, what the brand can credibly promise, what the engineering team can realistically ship, and which trade-offs are worth making explicitly. It is where someone has to determine whether a proposed solution is merely plausible or actually right.
AI can generate plausible interfaces from prompts. What it cannot yet reliably do is determine whether a proposed experience addresses the correct problem, supports a coherent strategy, or creates value that holds up once the product leaves the design canvas and meets actual users. It produces outputs. It does not own judgment, and it does not bear the consequences of being wrong.
AI is not compressing product design. It is compressing one layer of execution that many people mistakenly believed was the whole job.
The real question is not whether AI can generate UI. It can, and it will continue to improve. The more important question is whether generating UI was ever the heart of product design to begin with. It was not, and the sooner that distinction becomes broadly understood, the more useful this conversation becomes.
None of this means that AI-generated design is trivial or unimportant. It is neither. The practice that has come to be called "vibe design"—describing what you want, generating something plausible, and iterating until the direction feels right—is fast, often surprisingly capable, and genuinely useful in the right context. Early-stage exploration. Rapid concepting. Making an abstract idea visible enough to discuss.
The trouble is that products are not experienced in isolation. They exist within systems of logic, flows, states, edge cases, governance, brand constraints, technical limitations, and the accumulated decisions that came before. That is where vibe design begins to fail.
The first failure mode is context. AI can respond to prompts with impressive fluency, but it does not understand the business situation behind the decision. It does not feel the strategic tension between what a company wants to do, what it can afford to do, and what it must not compromise under any circumstances. A screen can look entirely convincing in isolation and be completely wrong in context, and recognizing that difference requires knowledge that no prompt can supply.
The second failure mode is coherence. Generating one compelling screen is not the same as sustaining consistency across a living product. Once the work expands across flows, states, exception handling, interconnected features, and the long arc of product evolution, AI-generated outputs tend to drift. Patterns become inconsistent. Logic fractures. The cleanup costs accumulate in ways that erode whatever efficiency was gained in the initial generation.
The third failure mode is operational quality. AI-generated design files are frequently bloated, over-structured, and architecturally messy in ways that carry real downstream costs: slower handoff, more engineering rework, and design systems absorbing quiet debt. What feels efficient in exploration often becomes expensive in production.
Vibe design is a useful shortcut to a direction. It is far less reliable as a substitute for the rigor required to build coherent, scalable products under the pressure of real-world constraints.
None of this is an argument against the tools. In fact, quite the opposite. AI is already changing design work in meaningful ways, and the honest assessment is that most of those changes are positive.
The areas where it creates genuine value are the areas that surround judgment rather than replace it. It helps synthesize research more quickly and surface patterns in qualitative data that would otherwise take days to identify. It accelerates exploration, compresses the distance between an idea and something concrete enough to react to, and reduces the particular friction of early inertia: the blank canvas, the first draft, the first direction worth arguing about.
Much of design work has always involved necessary but lower-leverage labor: organizing insights, clustering themes, generating initial options, and preparing starting points for iteration. AI can compress that work dramatically, freeing up time for the decisions that actually shape outcomes.
When anyone can generate ten screens, the scarce skill is no longer producing options. It is knowing which option is worth pursuing—and being able to defend that judgment under pressure.
The irony is that this amplification raises the stakes on judgment rather than diminishing them. Faster iteration creates faster feedback loops. More options increase the cost of making a poor choice. If generating screens is no longer scarce, the value of knowing which screen is right and being able to articulate why compounds considerably.
AI does not eliminate the need for designers. It elevates the importance of the part of design that has always mattered most.
What is actually happening to product design is not disappearance. It is a re-sorting: a clarification of where value truly sits inside the profession, and a widening gap between the work that AI is making cheaper and the work it is making more important.
The fundamentals remain essential. Human-centered thinking, interaction design, systems thinking, visual execution, product reasoning. These are not made obsolete by AI. They are what allow a designer to evaluate quality in the first place. They train perception. They make discernment possible. Without them, a designer cannot distinguish good output from merely plausible output, which is precisely the distinction that matters most in an environment full of both.
But in a world where generic execution is increasingly cheap and accessible, fluency in the fundamentals becomes table stakes. It is necessary. It is no longer sufficient.
One genuine area of differentiation is judgment and taste—not taste as an aesthetic preference, but taste as a form of discernment: knowing what is worth emphasizing, what should be simplified, what should be removed entirely, and what makes the difference between a product that feels purposeful and one that merely functions. As AI produces abundant, competent-looking output, that kind of discrimination becomes considerably more valuable.
Another is strategic depth. Designers who can connect their decisions to business logic, who understand growth levers, market positioning, and operating realities well enough to participate in those conversations credibly—will find themselves more useful as purely executional work becomes easier to automate. The ability to explain not just what was designed, but why it serves the business, remains uncommon. It is becoming more important.
A third is technical depth. Designers who can work fluently across systems, implementation constraints, and code occupy a position that is difficult to commoditize. As the boundary between design and development continues to dissolve, the people who can translate between the two disciplines become disproportionately useful to the organizations that need both.
The profession is not collapsing. It is broadening into a spectrum, with design fundamentals at the core and strategic depth, technical fluency, or sharper judgment becoming the dimensions along which the strongest practitioners distinguish themselves from the rest.
AI is not eliminating the need for product designers. It is eliminating the assumption that producing artifacts is enough to define the role.
As execution becomes cheaper, the market for generic output contracts. What becomes more valuable, and may become considerably more valuable, is the capacity to navigate ambiguity, make defensible decisions, and produce work that feels genuinely purposeful in a world increasingly crowded with output that does not. Products still require people who can turn uncertainty into direction, systems into experiences, and business intent into something users actually want to engage with.
That is not a threat to product design. It is a clarification of what product design was always supposed to be.