Writing
Essay · AI

When Software Starts Behaving, What We Design Changes

As software starts behaving, interface design is no longer the whole design problem. A reflection on why AI is changing what UX designers spend their time designing.

Mingcheng Guo · Apr 10, 2026 · 5 min read

Over the past year, I've read countless takes on what AI means for designers.

Dylan Field has argued that empathy and taste will become even more valuable. Jenny Wen talks about designers becoming increasingly T-shaped. Ryo Lu believes designers will eventually ship production-ready code. Others imagine a future where everyone becomes a builder, and the boundaries between design, product, and engineering begin to blur.

I find most of these ideas compelling.

But after following these conversations for a while, I noticed they're all trying to answer the same question: What will designers become?

What if the bigger shift isn't who designers become, but what we're actually designing?

A Familiar Conversation

That question didn't come from an article or a conference talk. It came from work.

A few years ago, we spent most of our time talking about information architecture, interaction flows, layouts, and discoverability.

Today, whether I was designing AI workflows at Rubrik or helping early-stage AI startups think through agent experiences, I noticed our conversations changing — the discussions sound more like this:

  • When should the AI take initiative?
  • What should it remember?
  • When should it explain its reasoning?
  • When should it wait?
  • When should it keep going on its own?

These aren't really interface questions.

Yet they've become some of the most important UX decisions I make.

It made me stop and ask:

What exactly are we designing now?

Software Changed

For a long time, software didn't really act on its own. People initiated actions, and software responded.

For most software, the interface was the behavior. That's where nearly all meaningful design decisions happened.

Then AI arrived.

Software can now remember previous interactions, understand context, plan ahead, reason through problems, and take initiative.

Software started behaving.

Which means we're increasingly designing software behavior:

  • Should the AI act or ask?
  • Should it explain or stay quiet?
  • Should it continue or pause for confirmation?
  • Should it remember this preference next time?

Sometimes it feels less like designing screens and more like designing the communication protocol between a person and an intelligent system.

Growing as a Designer

This shift has changed how I learn, too.

Lately, I've found myself reading far beyond traditional UX books.

Organizational behavior, communication, decision science, economics, even history.

Design discussions rarely have one correct answer. The perspectives you bring into a problem often shape the decisions you make more than any framework or methodology.

As software becomes an active participant instead of a passive tool, I've found those perspectives becoming increasingly valuable.

Why This Excites Me

One of my favorite classes at Carnegie Mellon was Brad Myers' Interaction Techniques (05-440/05-640 Interaction Techniques).

I still remember listening to his stories about Xerox PARC.

Windows.

Icons.

Direct manipulation.

Interaction techniques that feel completely ordinary today all had to be invented by someone.

A small group of designers and researchers fundamentally changed how billions of people would interact with computers.

We're still using the interaction language they created decades later.

Looking back, I think that's one of the reasons I became a designer.

I've always been drawn to the idea that careful observation, thoughtful design, and a few good decisions could genuinely change how people interact with technology—and in doing so, influence millions, or even billions, of people.

For a long time, I thought moments like Xerox PARC belonged to history.

Today, I think we're living through another one.

Maybe, a few years from now, conversations about agents, memory, planning, permission, and behavior will feel just as ordinary as windows, icons, and drag-and-drop do today.

If that's true, we're defining another generation of interaction.

When software starts behaving, what we design changes.

I'm grateful to be designing through that transition.

Comments 0