About three months ago, I opened a tool that promised to turn my Figma design into a live website. One click. No developer needed. No code.
I pressed the button. Something published. I shared the link with my team.
A developer came back twenty minutes later with a look I can only describe as exhausted sympathy. I won't repeat what was said. But the short version was: what went live had over a hundred things wrong with it that real users would hit immediately. Things I couldn't see. Things I didn't know to check for.
I had done everything right, as far as I knew. The tool had just quietly done everything wrong.
Here's the thing nobody is saying out loud
AI tools aimed at designers were marketed with one very specific promise. That you wouldn't need technical knowledge to use them. That the gap between "I designed this" and "this is live and working" was finally closed.
That promise turned out to require more technical knowledge than the old way, not less. You still need to know what a broken output looks like. You still need to understand enough to evaluate what the AI gave you. The tool does the doing. The judgment still has to come from you. And if you don't have the technical context to judge, you're flying blind.
A 2025 report from Figma found that 82% of developers were satisfied with AI features in their tools. Among designers, that number dropped to 54%. That gap is not a coincidence. Developers are using AI to speed up work they already understand. Designers are being handed AI and told to produce outcomes they've never been trained to evaluate.
What I actually experienced
I've been using AI tools in my design work for about a year now. Genuinely, some of it has been great. Renaming layers. Generating placeholder copy. Mood boarding faster. These things work.
But every time I tried to use AI for anything that would leave my hands and go somewhere real, a published page, a component someone else would build from, a file that would actually be used, something went sideways. Not in an obvious way. In a quiet way. The kind where it looks fine until it doesn't.
According to Nielsen Norman Group, the AI tools designers are actually adopting in their real work are narrow, focused on one specific task. The broader tools that promise to generate entire designs or whole websites are not in serious professional use.
That matches everything I've felt but couldn't name. The small, specific things work. The big, exciting promises don't, at least not without a technical safety net I was told I wouldn't need.
The honest, unresolved part
I don't know if this gets better, or how fast. I don't know if I need to learn more about how these things work under the surface, or whether that's even realistic alongside everything else I'm managing. I don't know if the tools will catch up to the promise, or if the promise was always a bit too clean.
What I do know is that "you don't need to know code" was never fully true. It was more like: you don't need to write code, but you still need to know enough to know when something is quietly broken.
That's a different skill. Nobody offered to teach it. They just handed us the tools and called it empowerment.
One thing to try this week
Next time an AI tool gives you an output, a layout, a component, a page, before you send it anywhere, ask someone with technical context to take one look. Not to do the work. Just to tell you if anything is quietly wrong. Notice what they catch that you couldn't see. That gap is where your actual learning is.
What has been your biggest "wait, this isn't as simple as they said" moment with an AI design tool? I want to know I'm not the only one.
Discussion
Loading comments...