AI Has No Taste. That's the Whole Point.
Gradient blob. Glassmorphism. Dark mode with purple accents.
You've seen this site a thousand times. It's a different company every time.
The tools got infinitely capable. The output got eerily uniform. And it's getting worse, fast.
The uniformity problem
Open any AI design tool and ask it to create a SaaS landing page. You'll get something that looks professional. Clean layout. Modern typography. A hero section with a gradient and a floating dashboard mockup. It'll look like it was made by someone who knows what they're doing.
Now do it again. And again. And again.
Every output is plausible. None of it is good. Not because the quality is low, but because the decisions aren't decisions at all. They're averages. The model learned what SaaS landing pages look like by ingesting thousands of them, and it gives you the mean. The centre of the distribution. The safest possible answer.
This is the same dynamic that's exposing thin-layer SaaS products across the industry. When the barrier to producing something adequate drops to zero, adequate stops being enough. The products that survive are the ones that made deliberate choices, not the ones that assembled plausible defaults.
Why AI can't tell you which one is good
AI generates a thousand variations of plausible. It can't tell you which one is good. That requires something the model doesn't have: a point of view about what this specific product, for this specific audience, in this specific market, actually needs.
It doesn't ask the questions a designer asks. Who's this for. What needs to convert. What's the one thing this screen has to do. What should it feel like. What should it deliberately not be.
It just answers. Beautifully.
That gap between answering and interrogating is where all the value lives. Figma shipped more AI features than any design tool in history, and most of them answer questions nobody asked. The features that save real time are the ones that automate the mechanical parts, not the decision-making parts.
Taste is the last competitive advantage
Every other advantage is being commoditised. You can generate layouts. You can generate copy. You can generate entire product experiences that look professional and function correctly.
What you can't generate is the instinct to look at all of that output and know which one is right. To know that this particular asymmetry works better than the balanced version. That the "wrong" typeface is actually the right one because it creates tension that holds attention. That the default should be refused because the default is what everyone else already chose.
That instinct is taste. And taste is still human.
This is why brand strategy has to come before visual design. Without a strategic foundation, every decision gets made on aesthetics alone, and aesthetics is exactly the domain where AI output is most convincing and most interchangeable. Strategy gives you criteria. Criteria give you taste in action.
Handcrafted doesn't mean hand-pushed pixels
Handcrafted design just became more valuable, not less. But handcrafted doesn't mean hand-pushed pixels. It means hand-picked decisions.
The question that changed the brief. The asymmetry left in. The "wrong" typeface chosen on purpose. The default refused. These are the things that separate a product from its competitors. Not the execution quality, which AI has already levelled, but the thinking quality, which AI hasn't touched.
I've shipped a working app with AI tools in a weekend. The code was solid. But the decisions about what to build, how it should work, what to leave out, those came from twenty years of designing products. AI removed the bottleneck between ideas and shipping, but it didn't remove the need for the ideas to be good.
The designers who get this are more productive than they've ever been. They're using AI to move faster on execution so they can spend more time on the decisions that actually matter. The ones who don't get it are using AI to skip the decisions entirely, and their work looks like everyone else's.
The template trap
If your product looks like every other product on the same template, it's going to get bucketed with every other product on the same template. Including the bad ones.
This was already true before AI. Now it's exponentially worse. When a founder can generate a complete product interface in an afternoon, the visual baseline goes up but the differentiation goes down. Founders already hire designers for the wrong reasons, and the temptation to skip design entirely when AI can produce something that looks good enough is stronger than ever.
But "looks good enough" is a trap. It looks good enough compared to what? Compared to the thousand other products that used the same tools with the same prompts to solve the same category of problem? That's not a competitive position. That's a crowd.
The $2.5 trillion being spent on AI is building better tools for generating plausible output. Almost none of it is going into helping teams make better decisions about what to build. The investment is entirely in the part that was already getting commoditised.
The next five years
The next five years belong to the teams whose work looks like it was made by someone, not generated by something.
That doesn't mean rejecting AI tools. It means using them in the right layer. Let AI handle the mechanical execution. Let humans handle the judgment calls. The brief. The constraints. The trade-offs. The moments where the right answer isn't the obvious one.
AI answers the brief. Designers interrogate the brief. That difference is everything.
The best design work has always been defined by the quality of thinking behind it, not the polish on the surface. AI just made that distinction impossible to ignore.
---
