Okay so I have been using AI tools as my primary design environment for the past two weeks. And I want to talk about what I actually observed, not the twitter version of it.
These tools work. In some places they work really well actually. First pass explorations, getting something on the screen fast, basic UI patterns that are already solved problems. Sure. But somewhere around day four or five of this experiment I started noticing something that bothered me. And it took me a while to put words to it.
The tool was making decisions without making decisions.
Like I would generate a screen and it would look reasonable. Competent even. But when I sat with it and asked myself, okay why is this the right pattern here, why is this the hierarchy we chose, why does this interaction work this way and not another way, there was nothing there. No answer. Just output. And in five years of doing this I have learned that the most dangerous design work is the kind that looks finished but has no thinking underneath it. Because that thing is going to get built. And six months later someone is going to sit in a room and ask why users are dropping off here and nobody is going to have an answer because the decision was never actually made. It was generated.
And here is the thing about speed, the thing that I keep coming back to. Speed is a very easy thing to sell. You can show speed. You can screenshot it. You can make a very satisfying gif of something generating in real time. What you cannot show in a gif is the forty minutes I spent last Tuesday just sitting with a single flow, no screen open, just thinking, because something felt wrong and I needed to find where it was wrong before I touched anything. That is not a slow moment in my process. That is the most important moment in my process. That is the moment where I earn my keep.
AI cannot do that. Not even close. And the tools are not going to close that gap by getting faster.
But here is where I think we are collectively getting it wrong and I include myself in this because I fell into it too. We are treating AI like a replacement for the process when it should be infrastructure around the process. And those are completely different things.
Think about what happened when designers moved from Photoshop to Figma. The designers who thrived were not the ones who knew Figma the best. They were the ones who understood design deeply enough that Figma just gave them a better environment for the thinking they were already doing. Components clicked because they already thought in systems. Auto layout clicked because they already understood how interfaces breathe and respond. The tool accelerated something real. It did not manufacture it.
What I have started doing, and this is genuinely where I think the leverage is, is mapping my actual process. Not the ideal version of my process that I would describe in a job interview. The real one. Where does time go. Where do I make decisions that require me and where do I do things that honestly do not require five years of experience. And then I ask, okay which of those second category things can I build a repeatable system around using AI. And I actually build the system. Document it. Refine it. Make it something that compounds over time.
Because here is what I have seen in five years of this work. The designers who consistently do the best work are not the ones with the best taste or even the most talent. They are the ones who built the best infrastructure around their thinking. The clearest process. The most intentional way of working. Everything else flows from that.
AI is not going to make you a better designer. But a better designer who builds the right system around AI, that person is going to be in a completely different category in three years. Not because of the tool. Because of the thinking they built around it.
The gap that opens up is not going to be between people who use AI and people who don't. It is going to be between people who built something real and people who are still sitting there prompting and calling it a workflow.