Most senior creative people will tell you they can spot AI work at a glance. Shallow, generic, a junior approximation of the real thing. I said a version of that myself. Then I spent serious time inside these systems and the belief didn’t survive. Some of the output is genuinely useful. Some of it is excellent. Some of it… I would’ve been proud to ship five years ago.
Which breaks a tool every experienced person in this field leans on: the read. A year ago, mediocre work still looked mediocre. Weak writing read as weak. Thin concepts looked thin. Generic interfaces looked unfinished. You could spot the gap between someone who had put in the reps and someone who hadn’t, often from across the room.
That gap is closing faster than I was ready for.
A single person with decent prompting can now produce work that looks “polished” in an afternoon. Design, writing, illustration, motion, code, strategy decks, product copy, research summaries, interfaces. The average output across every creative discipline is getting smoother and faster. In a field experiment Harvard ran with BCG, consultants using GPT-4 finished tasks 25% faster and their work was rated 40% higher in quality.
The floor is rising.
That changes the economics of creative work. It also raises a question I don’t think most of us have actually tried to answer: if a polished artifact no longer proves the person who made it is good, what does?
Before I answer that, it helps to look at what the artifact used to do for us.
For most of my career as a designer, creative work operated on visible effort. Production quality was how you proved yourself. Senior people had taste, juniors handled execution volume, and the ladder depended on the scarcity of craft. You could see who was good because their artifacts were good.
When execution becomes abundant, the artifact stops telling you what it used to. A polished interface no longer proves product judgment. A finished deck doesn’t prove strategic clarity, and confident writing doesn’t prove the thinking behind it was clear.
So that’s the answer. The problem isn’t quality anymore, it’s discernment. The real gap now lives between work that looks finished and work that actually solves the problem, and that gap is invisible to anyone who only judges by the surface (which, most of the time, is most people).
Coherence is not intelligence
I keep watching teams mistake coherence for intelligence in AI products. The interface feels smooth. The microcopy sounds confident. The interaction appears thoughtful. The empty states are charming. Then you use the product for five minutes and realize the system has no real grasp of what the user is trying to do.
The product is aesthetically complete and strategically hollow. This is a failure mode that barely existed before. If a team had the craft to ship something this polished, they probably also had the discipline to figure out what it was for. Polish was expensive enough to filter out confusion.
A recent example. I sat through a demo of an AI writing assistant that did everything right at the surface level. The onboarding was beautiful. The empty state had a small, friendly illustration. The first prompt suggestion was crisp. By prompt three, the model was hallucinating internal team names that didn’t exist. The team had spent months on the visual system and weeks on the model behavior. You could feel the asymmetry the moment you used it.
That asymmetry is what design leadership is becoming. It’s less about making the thing prettier, and more about catching the moment where the prettiness is doing rhetorical work the underlying system hasn’t earned.
What happens to apprenticeship
The part I think about most is what this does to how people become senior.
For years, junior roles in creative fields were built around repetition. You laid out a hundred pages and started to see why some breathe and some don’t. You wrote a hundred headlines and started to feel which ones land. You shipped a hundred features and started to notice which ones cause the support queue to spike on Tuesday. Production taught you pattern recognition, pattern recognition slowly built instinct, and instinct (given enough years of it) turned into taste.
The reps were physical when I started. 2005, art school: you laid out spreads for print, walked them to the output lab, and mounted the prints on foam core the night before crit. Spray adhesive, X-ACTO, steel ruler, 2am. One air bubble under the print and the whole room sees it at nine. You hand-coded your portfolio site in HTML. You animated your name in Macromedia Flash, frame by frame, and watched the timeline back a hundred times. Every tool in that list is dead now. The lessons aren’t, because none of them were really about the tool. Mounting a print badly teaches you what a sixteenth of an inch looks like. Watching your own Flash loop for the hundredth time teaches you exactly when a thing starts to bore you. It was discernment training dressed up as production work.
That’s how I got whatever eye I have. Foam core first, then hundreds of layouts at newsroom speed, most of them forgettable, and somewhere around the four hundredth I could feel a page going wrong before I could say why. Nobody assigned me that lesson. It was a side effect of the volume.
The reps are OPTIONAL now, which is the part that actually worries me.
A young designer can generate fifty landing page variations before lunch. A junior writer can synthesize five positioning directions in ten minutes. A product team can mock up an entire flow without opening a design file. The friction that used to do the teaching has been removed. You can already see it in the hiring data: a Stanford study found a 16% relative decline in employment for early-career workers in the most AI-exposed jobs, while more experienced workers in the same jobs held steady.
This is the question I don’t have a clean answer for. Taste still requires exposure. Judgment still requires friction. Good instincts still come from sitting with the discomfort of something not working and figuring out why. AI accelerates the output. It doesn’t automatically accelerate the part of you that knows when the output is wrong.
I think we’re about to live through a strange decade of people who can produce senior-looking work without ever having developed senior-level judgment, and the industry is going to spend a while figuring out the difference. The people who develop real discernment will do it on purpose, against the grain of how easy everything has become. They’ll take the friction the tools removed and add it back deliberately. Critiquing their own work harder, studying the things that didn’t work, and spending time with users instead of with finished artifacts.
I don’t think anyone’s designing apprenticeship programs around this yet. They probably should be.
Where seniority moves
If I had to name the defining senior skill of the next decade, it’s not craftsmanship, and it’s not prompting or software fluency either. It’s judgment.
The ability to look at ten plausible directions and recognize the one that actually works. The ability to hold human trust, narrative, pacing, ambiguity, interface behavior, emotional response, organizational reality, business incentives, and system constraints in your head at the same time when you’re deciding what to ship.
That’s harder than execution. It always was… execution was just easier to measure.
The future creative organization, the one I want to work inside, is smaller and higher output, with fewer production bottlenecks. But heavier at the top, because the real value now sits with people who can shape intent before the machine generates the artifact. People who define quality up front instead of inspecting it on the way out. People who understand that the interface is no longer static, that system behavior is the product now, and that designing system behavior requires a different kind of authorship than designing screens.
The floor will keep rising. That’s the easy prediction. The harder one is that the ceiling will rise too, but only for the people who treat judgment as a craft and practice it like one. Everyone else will produce competent, finished, hollow work and wonder why it isn’t landing.
For me, this is the most interesting moment my field has had in a long time. The tools got bigger. The job got smaller and more interesting at the same time. The thing that always mattered most, knowing what to make and why, is finally the thing that matters most visibly.
That’s worth raising my own standards for.