AI Design Systems made of cardboard.

Let’s be clear. In many of these cases, what’s being sold as a “design system” is, in reality: a collection of styles, four basic components, a lot of enthusiasm—and very little experience with real-world projects.

What we used to call “properly configuring a framework” now apparently needs a more serious name. More strategic. More… marketable. Because of course, it’s not the same to say: “I’ve set up a coherent theme”;  as it is to say: “I’ve created an AI-powered Design System”. The second one sounds much more… cool.

The logic is the same as always: take something simple and wrap it in enough complexity to make it look important.

If this feels familiar, it’s not a coincidence. It’s exactly the same pattern we saw with workshops: simple projects—corporate websites, products with little real complexity—going through oversized processes to “discover” things that were obvious from the start.
Now we’ve just replaced the meeting room with AI. But the logic is identical: take something simple and wrap it in enough complexity to make it look important.

Don’t get me wrong: AI is useful. Very useful. It lets you build a visual base in minutes, define tokens, generate components, and document everything with little effort. Perfect. The problem starts when we confuse that with having built a system. Because a design system doesn’t appear when you define colors.

A design system becomes necessary when you have real problems:

• Consistency across teams (Do you even have multiple design teams?).
• Conflicting decisions (Are there so many stakeholders that you’re losing your mind?).
• Edge cases that don’t fit (Is your product actually going to scale in 6 months?).

So—what problem are you actually solving by creating a design system? If you have to think too hard… there probably isn’t one.

These are not things you solve with a prompt and the test is pretty simple.
Take the most typical case: a corporate website. Everything is clear. Everything is defined. Low complexity. Zero scalability.
Quick question: What problem are you solving by creating a design system? If you need to overthink it… there probably isn’t one.
And then what always happens, happens. You start with something simple and end up with: cryptic token naming, overly abstract components, rules designed for a future that never comes.
The result: another small digital Frankenstein. This time, powered by AI.

Why do we keep falling for this? Because it sells.

Complexity looks professional. AI looks innovative. And “design system” sounds important.
Common sense, on the other hand, doesn’t generate much engagement.

But the reality hasn’t changed: not every project needs a design system and not everything that looks like a system… actually is one.

Sometimes, doing the simple thing well is enough. No theater. No inflated labels. No building, once again, another beautiful cardboard design system.