Everyone Can Generate a UI Now. So What?

A few days ago, I attended "Process Makes Perfect," a discussion on design process organized by Singapore Product Design and DBS. The room was full of designers, product people, and engineers. The kind of crowd that actually ships things.

One of the presenters, Pizza Yap, put up a slide on Motion & Interaction that hit the room differently than anything else that evening. The headline: "AI can generate UI. It can't yet make things feel good." And then, in bold: "If you want to stand out in a world where everyone can generate a screen, learn to make that screen feel alive."

I have not stopped thinking about it since.

Because he is right. And I see it every day in my work at SAP, building AI experiences. You can prompt your way to a dashboard, a settings page, or a checkout flow. Seconds later, you get something back. Good hierarchy. Clean spacing. Components that feel familiar because they are. The model has seen a million of them.

But then you click around. Navigate between views. Scroll. Try to actually use the thing.

Something is off.

It works. Technically, nothing is broken. But it does not feel like anything. It is flat. Every interaction is instant in the worst way. No weight. No response. No life behind the screen.

Pizza nailed it. AI is getting scarily good at generating interfaces and absolutely terrible at making them feel good to use.

That gap is where we come in.

The layout problem is basically solved

Let us be real. AI is impressive at arranging things on a screen. It knows primary actions should be prominent. It knows whitespace creates breathing room. It knows how a sidebar should relate to a content area. Feed it a description, and you get a reasonable layout in seconds.

This used to take hours. Sometimes days. Now it takes a prompt. For product builders, this is a genuine unlock. I use it myself. Why would I not?

But here is the thing nobody wants to say out loud: layouts were never the hard part. They were the time-consuming part. The actual hard part was always what happens after the layout exists. How does it feel when someone uses it?

That part has not gotten any easier. If anything, it has gotten harder to prioritize because the fast stuff got so much faster.

What "feels good" actually means

When I say an interface feels good, I am not being vague. I am talking about very specific things.

The toggle that does not just snap from off to on but slides with a slight overshoot before settling, like it has actual weight. The list item you deleted does not just disappear but collapses, and the items below it flow up to fill the space. Your brain registers that the world just reorganized itself because of something you did. That matters.

The card that lifts slightly when you hover, telling your hand it found something before your conscious mind catches up. The pull-to-refresh that gives you resistance, like you are stretching something real, and then snaps back. The page transition that does not hard-cut between views but slides or morphs in a way that tells you where you are in the product.

None of this is decoration. It is the difference between software people tolerate and software people actually enjoy using. And in enterprise, where users are stuck with whatever tool we give them, this distinction matters even more. If they cannot choose to leave, the least we can do is make staying not feel like a chore.

Why AI cannot do this

I have tried. I have prompted AI to add animations, transitions, and micro-interactions. It can technically generate the code. A fade here, a slide there.

But it does not feel right. And the reasons are not just "the technology needs more time." The gap is structural.

Feeling lives between states, not inside them. A layout is a single frame. AI is great at generating frames. But feeling is what happens when you move from one frame to the next. It is temporal. It unfolds over 200 milliseconds. AI has no concept of what those 200 milliseconds should feel like. It is guessing based on patterns it has seen, not based on anything it has experienced.

Good feeling requires deliberate imperfection. The things that make interfaces feel alive are technically "wrong." A button that overshoots before settling. A scroll that bounces past the edge. A progress bar that speeds up near the end so it feels faster even though it is the same speed. These are human tricks. We break the rules on purpose because we know how physics and motion feel in our bodies. AI converges on the correct pattern. Aliveness comes from knowing where to break it.

AI has no body. This sounds philosophical. It is not. When I decide a button press should feel "heavy" or a swipe should have "momentum," I am drawing on a lifetime of touching actual things. I know what resistance feels like. I know what weight feels like. I map those feelings onto pixels. AI does not have hands. It does not know what heavy means. It can pattern-match the word to some animation curve it has seen labeled that way, but it cannot feel whether the result is right.

AI has no opinion about who your product is. Open Linear. Then open Notion. Both well-designed. They feel completely different. Linear feels fast, precise, almost aggressive. Notion feels warm, a little playful, hand-crafted. That difference is almost entirely in the interaction layer. Timing, easing, how things enter and leave. These are personality decisions. AI does not know what your product's personality should be. It cannot decide if your interface should feel like a scalpel or a paintbrush.

This is our job now

I think a lot of product builders and designers are anxious about AI making them irrelevant. I get it. But when I look at what AI actually produces, the anxiety feels misplaced.

AI took over the part of our work that was time-consuming but not hard. The layout. The component selection. The information architecture. These were important, but they were also the most mechanical parts of what we do.

What is left is the genuinely hard, genuinely human stuff.

You are the one who decides that navigating from a list to a detail view should feel like zooming in rather than cutting to a new page. That decision shapes how users understand your product at a subconscious level.

You are the one who decides that deleting something should feel weighty. Not casual. Not instant. A brief pause, a considered animation, a confirmation that this action had consequences.

You are the one who sets the tempo. A 150ms transition feels snappy and responsive. A 300ms transition feels smooth and deliberate. A 500ms transition feels cinematic. Same animation, completely different emotional register. You pick the one that matches who your product is.

You are the one who designs the moments nobody notices until they are gone. The hover state that signals interactivity. The skeleton screen that shimmers instead of showing a blank void. The error shake mirrors the way a person might shake their head.

This is craft. You cannot prompt your way to it.

So what do we actually do with this?

Let AI do what it is good at. Generate the layout. Pick the components. Get the skeleton up fast. I do this all the time, and it saves me hours.

Then do the work that actually matters. The work that turns a functional interface into something that feels alive.

Honestly, the best thing that could happen to our industry right now is for everyone to realize that AI solved the boring part. The interesting part, the part that makes users feel something, is still wide open.

That is the work worth doing. And it is ours.