AI can now make a product interface look credible faster than many teams can schedule the kickoff meeting.
That is not a small thing.
It can generate dashboards, landing pages, product flows, image directions, coded prototypes, marketing copy, and design variations in minutes. It can turn a static mockup into a working front-end experiment. It can take hours of production drag off the table. It can help a designer explore ten directions while the old process was still warming up.
I have no interest in pretending that power is imaginary.
AI is useful. AI is fast. AI belongs in the modern creative tool belt. Any designer pretending otherwise is choosing nostalgia over the tools in front of them.
The question is control.
There is a difference between letting AI do the job and letting a designer do the job with AI as a tool.
When AI does the job, the tool becomes the taste-maker.
When a designer uses AI, the designer stays responsible for the vision, the critique, the restraint, the style, the final call, and the standard the work has to meet before it deserves a name.
That distinction is about to matter more than almost anything else in design.
Good enough is now the floor
A weak critique says AI only produces bad design. That critique is already outdated.
The more useful concern is that AI can now produce acceptable design at scale.
It can make something look complete. It can follow familiar UI patterns: readable cards, soft shadows, clean type, standard dashboards, neutral palettes, safe navigation, common onboarding flows, and layouts that feel like they belong somewhere in the modern software world.
Useful, yes. Finished, no.
A 2026 study, Usable but Conventional, found that AI-generated interface prototypes scored well on pragmatic UX qualities like usability and efficiency, while scoring weaker on originality and innovation.
That feels exactly right.
AI is very good at reaching the visible baseline. It can make the thing usable. It can make the thing look plausible. It can get the work to the point where a busy team might be tempted to stop.
But usable is not the ceiling.
Good enough used to be a milestone. Now it is a default setting.
If everyone can generate a polished, functional, familiar product surface, then polished, functional, and familiar no longer separate the work. They only prove that the tool was available.
The designer becomes more important here, not less.
When the floor rises, the ceiling matters more. Taste matters more. Judgment matters more. The ability to look at something that technically works and still say, "This is not alive yet," matters more.
AI can generate good enough. A designer pushes past it.
The tools are proving the point
The AI design and build market is moving quickly: Figma Make, Figma Sites, v0, Lovable, Framer AI, Adobe Firefly, Cursor, and a growing class of prompt-to-product tools are shrinking the distance between idea, prototype, code, content, and published work.
The headline is speed.
The better story is control.
The stronger tools are not simply giving people a blank prompt box and asking them to trust the machine. They are adding planning modes, design context, visual editing, point-and-edit controls, version history, Figma imports, code review, annotations, comments, style controls, and refinement loops.
Figma Make frames the work as code-backed and visually editable, with context, annotations, point-and-edit controls, and version history inside the process. Framer frames its AI workflow around generation and refinement while keeping the user in control. v0 and Lovable promise fast app generation, but their workflows still include reviewing, iterating, syncing, inspecting, and refining.
The product language gives the game away.
Generation is not enough.
Somebody still has to know what good means.
The tool can produce options. The designer decides which option deserves to live. The tool can speed up production. The designer decides whether speed has created value or only created more output. The tool can generate a surface. The designer owns the standard.
AI belongs in the tool belt. It does not replace the hand, the eye, or the mind of the artist.
The problem is superficial competence
We need sharper language than "AI slop," even though the term points at something real.
The sharper problem is superficial competence: work that looks responsible before anyone has really tested its judgment.
It borrows the appearance of care.
You can feel it in the tells: safe layouts, muted palettes, beige or tinted backgrounds, blue and purple gradients, oversized rounded rectangles, predictable sans-serif type, generic dashboard cards, drop shadows with no hierarchy, robotic product copy, and error states that sound like placeholder text.
The product can look expensive at a glance and feel empty after thirty seconds.
Business Insider recently described a similar pattern in AI-coded apps as a "regression to the mean": polished products converging around the same visual habits. The academic paper Why Slop Matters gives the stronger phrase: superficial competence.
That phrase matters because the danger is not always ugliness. Sometimes the danger is polish.
Superficial competence looks finished enough to avoid being questioned. It gives a team permission to move on before the design has been pushed, sharpened, contradicted, or given a point of view.
The same tension is showing up outside product design. Across visual culture, there is already a reaction against synthetic sameness and frictionless machine polish. People are reaching again for work that feels tactile, authored, imperfect, specific, and alive.
When machine polish becomes abundant, human authorship becomes a premium signal.
The answer is not to reject AI.
The answer is to reject default output.
AI doing the job vs. a designer using AI
Creative control is the center of this argument.
When AI does the job, the first or second polished draft gets treated as finished. Generic patterns are accepted because they look safe. Speed gets mistaken for design quality. The human reviews the output instead of directing the work.
The final product carries the taste of the tool.
When a designer uses AI, the process changes.
The designer defines the desired result before generation. The designer knows the audience, the feeling, the hierarchy, the constraints, the brand, the level of delight, and the reason the thing should exist.
AI helps explore options quickly. The human critiques, curates, rejects, redirects, and refines. The designer maintains authorship.
AI can be an assistant, a collaborator, a production tool, a sketch partner, a critic, and an execution layer. It can research, organize, summarize, code, test, structure, and help produce the work around the work.
But it should not own the vision.
A designer is not there to babysit output. A designer is there to own the standard.
That means being present before generation, during iteration, and at final judgment. It means feeding the system better inputs, building repeatable processes, creating constraints, testing directions, and refusing to let the machine's first confident answer become the creative decision.
Raw generation is not directed craft.
Good UX is not the ceiling
A lot of conventional product UX is mature now.
Forms, dashboards, filters, sidebars, navigation patterns, onboarding flows, table views, settings pages, checkout patterns, modal systems, profile screens, empty states, and common interaction models have been studied and repeated for years.
AI can learn those patterns because the industry has made them visible.
So yes, AI can often assemble a usable flow.
Product design is still unsolved.
A working solution helps the user complete the task. A great solution makes the user trust the product, understand the system, feel the brand, enjoy the interaction, and remember the experience.
A working solution answers the functional question. A great solution also carries rhythm, confidence, clarity, restraint, surprise, and personality.
Design is harder to grade than code.
Code has clearer validation loops. It can compile or fail. Tests can pass or fail. Errors can be thrown. Performance can be measured. Behavior can be observed.
Code is not simple, and expert developers still matter. Research on AI-generated code has shown that code can be functionally correct while still carrying deeper issues around maintainability, responsiveness, accessibility, duplication, complexity, and long-term quality.
Code at least has more mechanical feedback loops.
Design has to be judged.
The compiler can tell you when code breaks. It cannot tell you when a product feels generic.
OpenAI's Andrew Ambrosino made a similar point in a Business Insider piece about why design is harder for AI to grade than software. The hard question is whether something works with taste, context, intent, and judgment.
That connects to my own relationship with AI and code.
AI now lets me push design work closer to a functioning product. I can create more than a clickable prototype. I can explore coded interfaces, test behavior, and build a more complete product design package than I could before.
But I am not a senior engineer. I am not always qualified to judge code quality at the level a proper developer can judge it.
That honesty strengthens the argument.
AI gives specialists more reach, but it does not erase expertise. A designer using AI can carry a product idea farther. A developer using AI can do the same in code. The best work still respects the difference between output and judgment.
AI may get the interface to usable. A designer has to get it to memorable.
Design is art put to work
I have heard people argue that art and design are distinctly different.
I understand the distinction. I do not accept the wall.
Art and design are not separate continents. They are overlapping currents. Sometimes expression is more visible. Sometimes function is more visible. Purpose does not cancel beauty. Function does not erase expression.
Nature proves this constantly.
A leopard's spots are designed for camouflage, but they are also beautiful.
The colors and patterns of a tropical bird may help attract a mate, but they are still visually stunning.
A bridge has to carry weight, but its curves, proportions, and silhouette can make it unforgettable. A well-crafted knife has to cut, but it can also be admired for balance, material, and form.
A Lamborghini Countach is purposefully designed as a vehicle. No one would deny the artistry in its shape.
Even chess follows rules and has an objective, yet a brilliant combination can feel like poetry.
The same is true in product design.
An interface can be functional and still carry life. A dashboard can be clear and still have rhythm. A workflow can be efficient and still feel like it belongs to a particular brand, product, and moment.
A product can solve a problem and still have soul.
Design is not the absence of art.
Design is often art under pressure.
Left alone, AI often satisfies the structure and misses the charge. It can produce the frame. It can solve the obvious problem.
But it does not always know when the work needs an unusual proportion, a stronger line, a quieter button, a more human sentence, a sharper hierarchy, a playful moment, a little restraint, or a little danger.
That is taste.
Taste is not decoration. Taste is judgment made visible.
The AI-enhanced creative director
The best use of AI is to protect the creative's highest-value time.
I want AI to take tedious work off my plate. I want it to research, organize, schedule, code, summarize, test, produce, refine, and help execute.
More than that, I want it to help me see more possibilities so I can make better decisions.
My division of labor is clear: I am the creative. AI can handle practically everything else.
That does not mean I disappear from the process. It means I stay where I am most valuable.
I maintain control over the decisions. I set direction. I define the outcome. I decide what the work should feel like. I decide when the first draft is not enough. I decide when the machine has produced something technically acceptable but creatively empty.
Repeatable AI workflows matter here. A serious designer should not treat AI like a slot machine.
Prompt, pray, accept.
That is not a process.
The process has to become structured: input, generation, critique, curation, iteration, craft, final judgment. Over time, those processes can become skills, systems, and reusable ways of working. The more capable the tools become, the more important the creative operating system becomes.
AI gives the designer reach, not permission to stop caring.
The goal is not to automate taste. The goal is to protect creative time so taste can do its work.
Who gets replaced and who gets stronger
There is a blunt version of this conversation: AI will replace junior designers.
That is too simple.
The real risk is not a job title. The risk is an outcome level.
A designer who settles for junior-level outcomes is vulnerable, regardless of seniority. A designer who treats good enough as finished is vulnerable. A designer who uses AI to generate a polished first pass and then stops is vulnerable.
AI can already produce a lot of that work faster.
But designers who integrate AI into a disciplined process become more valuable. They use it to expand exploration, test more directions, build stronger systems, sharpen their taste, and push beyond the obvious.
They move faster. They see farther. They spend less time dragging pixels through tedious tasks and more time making the decisions that actually shape the work.
In an age of infinite generation, taste becomes a survival skill.
The people most exposed by AI are not the people learning tools. They are the people defending mediocrity.
The warning and the invitation
Teams that rely on default AI-generated design will create products that feel increasingly interchangeable.
They may look clean. They may function. They may impress people for a moment. But they will feel like they came from the same average place, shaped by the same average assumptions, built from the same average patterns.
That is not where great design lives.
Use AI to get creative time back. Use it to explore farther. Use it to remove drag. Use it to build better systems. Use it to create more, not to care less.
Do not surrender taste, style, creativity, or the personal artistic touch that gives the work its personality.
AI should remove friction from creation, not remove the creator from the work.
If we give up our creative spirit to the machine, we risk producing design that gets the job done but lacks the life that makes the work worth caring about.
So yes, use the tools. Use them aggressively. Use them intelligently. Build processes around them. Let them widen the field. Let them shorten the distance between idea and execution.
But stay in control.
AI can generate good enough.
That is exactly why the designer has to push past it.