I’m going to say something that might be unpopular in a design community that’s currently falling over itself to prove AI fluency: most people are using AI wrong. Not because the tools are bad, but because the habits are.
This isn’t a screed against AI. I use it regularly and deliberately. But I’ve made some of these mistakes myself, I’ve watched others make them, and I think the design industry needs a more honest conversation about where things are going sideways.
Here’s what I’ve observed.
Mistake #1: Offloading understanding to AI
This is the root of everything else on this list. And if you only take one thing from this article, make it this.
Design is fundamentally an act of understanding.
Before a single decision gets made — before a frame is opened, before a component is named, before a single user flow is sketched — a designer’s job is to understand. The users. The product. The domain it exists in. The business goals. The technical constraints. The stakeholders and what they’re actually afraid of. The clients and what they’re not saying out loud.
That understanding is not transferable. It cannot be summarized into a prompt and handed off. And yet I see designers doing exactly that — feeding a brief into an AI tool and treating the output as a foundation to build on, rather than doing the slow, unglamorous work of developing genuine comprehension themselves.
The problem isn’t that AI gives bad answers. Sometimes it gives quite good ones. The problem is that a designer who doesn’t deeply understand what they’re building cannot evaluate whether those answers are right. They have no filter. They have no instinct to push back when something is subtly wrong. They’re building on a foundation they don’t own — and it shows in the work.
Understanding is not a phase you can skip and make up for later with better tooling.
Understanding is the job.
Mistake #2: Letting AI drive the build before the design is resolved
I’ll be personal here, because I think honesty is more useful than authority posturing.
Early in building out my own portfolio site, I made the mistake of letting AI take the reins on the development side before my design system was fully resolved. I was moving fast, the code was coming together quickly, and it felt productive. It wasn’t.
What I ended up with was a codebase full of hard-coded values that should have been variables. Colors defined inline instead of referencing tokens. Spacing values duplicated across files instead of pulled from a consistent scale. The design system I had so carefully architected in Figma had not made it into the code — not really. What made it into the code was a series of one-off decisions made in the moment by a tool that had no awareness of the system I was trying to build.
In OOUX terms — the Object-Oriented UX framework I work within — this is how you end up with shapeshifters, broken objects, and masked objects in your UI. A shapeshifter is an object that behaves inconsistently across your product, changing its attributes or actions depending on context in ways that confuse users. A broken object is one that should allow the user to take action, but doesn’t (missing CTA’s or certain attributes). A masked object one that has identical representation (cards, details) to one or more other objects’ representation, making it invisible or inaccessible to users who need it. All of these things contribute to poor user experience.
These aren’t just UX problems. They’re symptoms of a design system that lost coherence during implementation. And AI-assisted development, when it outruns design thinking, is one of the fastest ways to get there.
The fix isn’t to stop using AI…it’s to stay in the driver’s seat. Review every line. Catch the hard-coded values before they multiply. Don’t let the speed of generation outpace your ability to evaluate what’s being generated against the system you actually intended to build.
Mistake #3: Treating AI output as a first draft instead of a starting point
There’s a meaningful difference between a first draft and a starting point, and it matters more in design than almost anywhere else.
A first draft implies the structure is roughly right and needs refinement. A starting point implies you’re going to interrogate it, challenge it, and potentially throw most of it away. AI output — whether it’s a UI layout, a content hierarchy, a color palette, or a component structure — is a starting point. Always.
Designers who treat it as a first draft skip the interrogation step. They refine instead of question. And the result is work that feels slightly off in ways that are hard to articulate — because the foundational decisions were never examined, just inherited.
Mistake #4: Using AI to validate instead of to challenge
This one is subtle and worth paying attention to.
AI tools are, by nature, agreeable. Ask one if your design direction is good and it will find ways to affirm it. Ask it to review your information architecture and it will note strengths before weaknesses. This is not malice — it’s a pattern baked into how these models are trained to respond.
The designers getting the most value from AI are the ones who have learned to use it adversarially. Ask it to poke holes, not pat you on the back. Ask it what a skeptical stakeholder would say. Ask it what you’re missing. Ask it to make the case against your decision. That’s where it earns its place in a serious design practice.
Mistake #5: Confusing AI fluency with design maturity
This might be the one that matters most for where the industry is headed.
There is a generation of designers entering the field right now who are extraordinarily fluent with AI tools and significantly less fluent with the fundamentals those tools are supposed to augment. They can generate a UI in minutes. They struggle to explain why it works or defend it under pressure. They’ve learned to produce before they’ve learned to think.
AI fluency is a real and valuable skill. It is not a substitute for design maturity — for the years of developing taste, building systems thinking, learning to ask the right questions, and earning the instincts that let you know when something is wrong before you can articulate why.
The designers who will matter most in the next decade are the ones who bring both. Who can move fast and know what they’re doing. Who use AI as leverage on top of genuine expertise rather than as a shortcut around developing it.
That’s a higher bar than it sounds like. It’s also the only bar worth clearing.
The tool is not the problem. The thinking is always the problem. AI just makes it easier to move fast enough that you don’t notice the thinking was missing until it’s too late.
