.png)
When anyone can generate a polished interface, the real design work starts before the screen.
For years, producing the interface was one of the most visible demonstrations of a designer’s value. We translated requirements into flows, explored layouts, built components, refined interactions and eventually arrived at something tangible enough to put in front of a user. That work still matters, but the economics of producing it are changing quickly as generative tools make it possible to move from an idea to a convincing interface in a fraction of the time.
The interesting question is no longer whether AI can produce design work because it clearly can. The better question is what becomes more valuable when producing the first version is no longer particularly difficult.
Figma put the distinction plainly in its 2026 AI report: “AI can build almost anything. What it can’t do is determine what’s worth building.” Its research, based on more than 8,000 survey responses collected over three years, also found that 90 percent of respondents consider design at least as important as it was before AI, while nearly six in ten believe it has become more important. That might sound counterintuitive until you separate the act of producing an interface from the decisions required to create a useful product.
Product teams rarely begin with perfectly articulated problems waiting to be turned into screens. They begin with declining metrics, customer complaints, operational inefficiencies, stakeholder requests and assumptions about what users need, and someone still has to determine which of those signals represents a problem worth solving.
A request for a new feature may reveal a broken process, an information gap or an issue that does not need another feature at all. Generating the requested interface more quickly does not make the original assumption more accurate, which is why research and problem framing become more important as production gets easier.
This is also where I think designers have an opportunity to resist becoming overly focused on output. If our contribution begins when someone asks for a screen, we have already surrendered a significant part of the design process. The more valuable position is earlier, when the team is still deciding what the problem is, whose problem it is and whether solving it will meaningfully improve anything.
AI also changes the volume of work a team can produce. When ten plausible directions can be generated in the time it once took to explore one or two, the constraint shifts from our ability to create possibilities to our ability to evaluate them.
Adobe's 2026 research into creative work found evidence of exactly this shift. After examining job postings, surveying more than 1,400 working creatives and studying thousands of people preparing to enter creative fields, Adobe found that creative labor is moving away from first-pass variation and toward selection, direction, communication and accountability. The researchers ultimately described the pattern as a redistribution of creative work rather than the replacement of creative workers.
That distinction feels particularly relevant to product design because a plausible interface is not necessarily a responsible recommendation. Designers still have to determine whether a direction reflects what was learned through research, whether it supports the intended business outcome and what happens when the experience leaves the ideal path. We have to consider accessibility, edge cases, downstream consequences and the tradeoffs hidden behind something that may look perfectly reasonable in a generated prototype.
More possibilities are useful, but they also create more opportunities to confidently choose the wrong one.
The growing importance of judgment also changes the role of critique. Nielsen Norman Group recently argued that critique is becoming a core design skill in the AI era because designers increasingly need to define the criteria by which generated experiences should be evaluated. Their argument is not simply that humans should look at AI output and decide whether they like it; rather, designers need to translate what they understand about users into a rigorous definition of what good performance actually looks like.
That definition has to come from somewhere. It comes from observed behavior, research, accessibility standards, product strategy, domain knowledge and an understanding of the circumstances in which someone will actually use the product. Without that context, an AI system can produce something polished without having any meaningful way to determine whether it is appropriate.
For designers, this creates an interesting expansion of craft. We are not only designing the experience but increasingly defining the principles and criteria that allow other people, systems and tools to produce experiences without losing sight of the user.
This distinction becomes even more important in enterprise environments, where a seemingly straightforward interface change can affect permissions, operational processes, reporting, downstream systems or the work of teams the designer may never see.
A generated prototype can make a solution appear deceptively complete because the visible interface arrives before the surrounding system has been fully understood. Designers who can think beyond the screen are able to ask what happens before and after an interaction, identify dependencies that are missing from the requirements and recognize when simplifying one person's experience creates additional work for someone else.
Those questions have always been part of good product design, but they become easier to overlook when producing something visually convincing requires so little time.
There is a tendency to frame conversations about AI and design around replacement, but I find that question less useful than asking how the work itself is changing. Adobe's research found that creative job postings actually increased by 8 percent during the six-month period it examined, while explicit demand for AI skills also increased. The evidence is still early and should not be treated as a prediction of the long-term labor market, but it complicates the idea that better generative tools automatically make creative expertise less relevant.
If anything, easier production raises the standard for what designers should contribute beyond production. Knowing how to operate the tools will matter, but so will knowing when their output is mediocre, when the prompt is based on a faulty assumption and when a team is moving quickly toward something that should never have been built.
The future value of design may therefore be less attached to how efficiently we can produce another screen and more closely tied to the judgment surrounding it. Researching the problem, understanding the system, defining what good looks like and deciding what deserves to ship are not steps that happen around the design work.
Increasingly, they are the design work.