AI Made Design Easier. It Didn’t Make It Good.

Nice Scissors. Have You Ever Actually Cut Hair?

Buying a fancy pair of scissors doesn’t make you a hairstylist.

It doesn’t mean you shouldn’t get a great haircut, though. It just means you find someone trained to use them well.

That’s the tension I feel as I watch designers, and non-designers, rush into AI tools with a “build first” mindset. Because y’all, we’ve actually been here before. More than once.

The Desktop Publishing Wave

Back in the 90s, desktop publishing put typesetting into anyone’s hands. Before that, laying out a document (choosing type, setting margins, controlling hierarchy) was a skilled trade. Suddenly anyone with a Mac and a copy of PageMaker could do it themselves.

The results were, famously, not great.

Comic Sans everywhere.

Papyrus showing up on things that had no business being anywhere near ancient Egypt.

Six fonts on one flyer.

Text stretched and squeezed to fit a box instead of the box being designed to fit the text.

I want to be clear: Comic Sans and Papyrus are not inherently evil. They have their place. A kid’s birthday party flyer, sure. A dentist’s office logo trying to look “friendly,” I guess. But an untrained eye doesn’t pick them for the right reasons. They pick them because they’re fun, or different, or the first thing that jumped out. And then they use them for everything, forever, with total confidence. That’s the tell.

The tools had democratized production. They hadn’t democratized judgment.

It took years, and a lot of typography education finding its way into mainstream design culture, before the average document stopped looking like a ransom note.

The WYSIWYG Wave

Fast forward to the early 2000s, and the same pattern showed up again with website builders. Suddenly anyone could drag, drop, and publish a website without knowing HTML. The barrier to building something dropped to almost nothing.

And once again, the results were often unusable. Confusing navigation. Inaccessible color contrast. Layouts that worked on one screen size and fell apart on every other. The tools made it possible to build. They didn’t make anyone able to judge whether what they’d built actually worked for the people using it.

It took the rise of UX as a discipline, and later design systems as a governance tool, to close that gap. Not everyone became a professional designer. But the field built guardrails, standards, and shared components that helped non-specialists produce better outcomes than raw talent alone ever could.

Now It’s AI’s Turn

I’m seeing a lot of designers, and plenty of non-designers, jump into AI tools with a “build first” mindset. And honestly, I get the excitement. These tools can produce something polished-looking in minutes: clean typography, a coherent color palette, a layout that looks finished.

But polish and quality are not the same thing.

Someone without a trained eye sees something that looks better than what they could make themselves, and it’s easy to mistake that for good. That disconnect is the real risk of this moment. Not that AI produces bad output, but that it produces confident looking output that nobody involved is actually equipped to evaluate.

I think of it like a door with a handle that says “push,” when the only way through is to pull. The design looks like it affords one thing. It actually requires another. A trained eye catches that gap immediately. An untrained one just feels confused, or worse, doesn’t notice at all. And neither does anyone downstream who’s relying on the product working the way it appears to.

And it’s not just the visual layer. The interaction patterns are off too.

A dropdown that should be a radio button.

A modal that opens when a tooltip would do.

A multi-step flow where one screen would work fine.

AI is generating interfaces that don’t match the mental models people have built up over twenty years of using the web, because those mental models aren’t really “logic,” they’re convention, and convention is exactly the kind of thing that’s hard to reverse-engineer from a prompt.

Here’s the part that gets me the most: someone without design training can usually tell that something feels a little off. They just can’t say why. They’ll click around, get a nagging sense that it’s clunky, and shrug it off as a personal problem, like they’re just not tech savvy enough, instead of what it actually is: the product asking them to do the work the design should have done for them. Mind the gap — that space between sensing something’s wrong and being able to name it is exactly where a trained eye earns its keep.

Lipstick on a Pig: The Latest Shade, Now Available

Designers who’ve been staring at this stuff for a while have started calling it “lipstick on a pig.” Harsh, but honestly, kind of fair. And the other word that keeps showing up is “slop,” which, funny enough, is also what you feed a pig. Somebody out there has a sense of humor about all this, even if it’s a dark one.

That phrase captures something important: this isn’t one problem, it’s three stacked on top of each other:

1 - The visual design is often lacking, inconsistent, or just plain wrong.

2 - The UX and interactions can be misleading or actively broken.

3 - And the accessibility issues are real, whether or not the thing technically passes an automated scanner. Something can sail through every automated accessibility test you throw at it and still be genuinely unusable for a real person trying to get something done.

That’s the part that stings. Someone pours real time and energy into shipping this AI slop, honestly believing it’s the greatest thing since sliced bread, only to get rudely awakened when actual users show up and it turns out to be a hot mess. Nobody wins there. Not the person who built it, not the people trying to use it, and definitely not the mission the product was supposed to serve in the first place.

What Actually Changed, and What Didn’t

Here’s the pattern across all three waves: the barrier to production keeps dropping. The barrier to judgment never does.

Desktop publishing didn’t make everyone a typographer.

WYSIWYG tools didn’t make everyone a UX designer.

AI isn’t going to make everyone a product designer either, no matter how good the output looks on the first pass.

What each wave actually needed wasn’t less access to the tools. It was more access to the judgment that makes the tools worth using well: education, standards, and often, someone trained to look at the result and say, clearly and specifically, here’s what’s actually wrong, and here’s why it matters.

The Real Question

If you’re using AI to build something right now (a website, an app, a product), the question worth asking isn’t “does this look good?”

It’s “would this hold up to someone who’s spent years learning what ‘good’ actually requires, for the person using it, not just for the person who made it?”

That gap between looking finished and actually working is where the next generation of design problems is quietly getting built. The tools got faster. The judgment gap didn’t close. It just got easier to miss.

Trained Designers Are More Valuable, Not Less

Here’s the part that should give trained designers some hope in the middle of all this AI anxiety: every wave of “anyone can build it now” has ended the same way. The tools got adopted, the slop piled up, and the people who actually knew what they were doing became more in demand, not less. Judgment doesn’t get automated away. It gets more valuable the more noise there is to sort through.

And a lot of the sorting looks the same every time, too. Someone ships their AI-built thing, feels great about it for a while, and then real users show up and the cracks start to show. Eventually, sheepishly, they seek out a skilled designer. Not to start over. Just to fix it. To take the AI slop and make it actually work.

There’s no shame in that, honestly. Y’all know how this story goes. Maybe it happened to you. More likely you know somebody, a sibling, a college roommate, that one friend from your book club, who watched a few YouTube tutorials, bought some real clippers, and decided they could handle cutting their own hair. Or worse, offered to “help out” a friend. It goes fine right up until it doesn’t, and suddenly there’s a chunk missing that was never part of the plan. Everybody’s got a story like this, or knows somebody who does. That’s when you call an actual stylist, a little sheepishly, and ask them to fix it. Not because the clippers were bad. Because owning them was never the same thing as knowing how to use them. That’s still a real skill, still worth paying for, and if anything, this whole AI moment is making that clearer, not less clear.

And look, not every haircut needs to be a luxury salon experience. Sometimes a SuperCuts is exactly right. Quick, cheap, good enough, done.

There’s an old saying about projects: fast, cheap, good, pick two. Haircuts follow the same rule, and so does AI-built design. There’s nothing wrong with picking fast and cheap, as long as it matches what you actually need.

The problem isn’t reaching for that option. It’s not knowing the difference, or worse, thinking you’re getting salon results when what you actually got needs a professional to salvage it. Know which one you’re choosing, and choose it on purpose.

In closing…

I know where a lot of designers are right now, watching AI spit out layouts in seconds, quietly wondering: am I still needed?

Yes, absolutely.

You’re needed to know the difference: the gap between “this technically works” and “this actually serves the person using it.”

Convincing a stakeholder of that gap is the harder part.

A few places to start: stop arguing taste, ask for evidence. Try “have we watched anyone try to use it?” Show, don’t tell. Put it in front of five real users and let the gap reveal itself.

Reframe speed as a phase, not a finish line. Fast gets you a first draft; the distance from there to shippable is exactly the terrain designers know how to navigate. That’s the job now. Not producing faster than the machine. Minding the gap it can’t see.

AI Made Design Easier. It Didn’t Make It Good. was originally published in Bootcamp on Medium, where people are continuing the conversation by highlighting and responding to this story.