If you're new to product design right now, the anxiety usually shows up as one question. Is this field closing just as I'm trying to get into it? AI is generating interfaces, writing copy, even shipping working prototypes from a prompt. It's a fair thing to worry about.
The honest answer is that the field isn't disappearing. Its center of gravity is moving. For the last decade, a lot of designers were judged mainly on screens, a clean interface, a usable flow, a tidy component library. That still counts. But AI has made the production of an average screen cheap, so the value of design is shifting to the parts a prompt can't do on its own, deciding what to build and why, and understanding how the thing actually behaves once it's live.
The interaction model changed, not just the tooling
Most software before AI was command-based. The user did the navigating, you clicked a button, filled a form, chose from a menu, and the interface simply exposed what was possible. AI software is closer to intent-based. You say what you want, the system interprets it, plans, and acts, sometimes generating the interface or the next step on the fly.
Take a coding agent as an example. The interesting part isn't that it writes code. It's that the product only works if you can see the diff, inspect the files it touched, run the tests, reject a change, and undo it. None of that is a visual layer bolted onto a smart model. It's the design. The moment software can act on its own, the interface's job stops being “show the options” and starts being “show what happened, and let the user correct it.” That's a genuinely different skill than laying out a settings screen, and it's one very few junior designers are being trained for yet, which is actually an opening if you get ahead of it now.
Judgment is now the scarce resource
The old path into design was often downstream. Someone else decided what to build, product wrote the requirements, and design turned that into flows and visuals. That path still exists, but it's thinner than it used to be, because AI has compressed the execution side of the job. What's left scarce is judgment. Which problem is actually worth solving. What the simplest useful version looks like. Where the system should stay predictable rather than clever. What the product's taste is, and what it should refuse to do.
This matters specifically if you don't have five years of shipped work behind you, because it means the gap between a junior and a senior designer is no longer mostly about polish. Two people can produce equally clean screens with the same AI tools. What separates them is whether they can say, convincingly, why this version and not the other three.
What's actually worth practicing
Taste is the first. AI can generate a hundred variations of a screen in an afternoon, and most of them will be forgettable. Taste is the ability to tell which one is alive and which one is generic, and it's built by looking closely at real work, not by producing more of your own.
Clarity is the second, and it matters more with AI in the loop than without it, because ambiguity multiplies fast. If a user can't tell what the system is doing or why, they stop trusting it, so clear state, a clear next action, and a clear way to recover from a mistake stop being nice-to-haves and become the difference between a product people keep using and one they quietly abandon.
Writing is the third, and it's underrated by people new to the field. AI products are language-heavy, empty states, confirmations, error messages, the small sentence that explains why the system did what it did. Bad writing makes an AI feature feel evasive or dumb. Good writing is what makes it feel calm.
Prototyping with real behavior is the fourth. A static mockup can't show what happens when the AI gets something wrong, or when a user rejects a suggestion three times in a row. Learning enough code, through tools like Codex, Claude Code or similar AI-assisted builders, to make a prototype that actually behaves, is quickly becoming a baseline skill rather than a specialization.
A more honest self-check than a checklist
When you look at an AI feature you use daily, can you explain in plain language what it's allowed to do on its own, and where it has to ask first? If you can't answer that about a product you already use, that's the gap to close before anything else.
Can you fill in, for a product idea you care about, who it's for, what painful moment it solves, and why someone would come back a second time? If any of those come out vague, the screens aren't the problem yet.
None of this requires a job title to start practicing. It requires picking a real AI product, watching it closely, and being able to say precisely what you'd change and why.
This is the version of the field I actually see forming. If you're early in this and it feels unsettled, that's probably because it is. The fundamentals didn't get easier though. If anything, AI is making it harder to fake them, which is good news if you're willing to put in the work most people are skipping.
A good chunk of that self-check, especially explaining your reasoning out loud, is exactly what an interview panel is going to test you on, so it's worth practicing before you're in the room.
If you're trying to turn this into an actual story you can tell in an interview, not just a set of beliefs about the field, that's the part withLyra can help with. Talk it through with our AI interviewer and see where your reasoning holds up and where it still needs work.
Back to all articles
