AI: The Death of of the Hand-off

DAte

May 26, 2026

Category

AI

Reading Time

~3

Handoff between design and engineering was always a workaround. A polite little ceremony built around the fact that designers and engineers often weren’t close enough to the same problem at the same time.

Design made the thing.

Engineering built the thing.

Then everyone acted surprised when the thing that shipped was technically correct but experientially off.

The hierarchy got flattened.

The empty state disappeared.

The edge case became a support ticket.

The flow worked, but only if you already understood the company’s internal logic. Which, spoiler alert, users usually do not.

For a while, handoff was annoying but survivable.

But today, AI makes it expensive. Because now we’re not just designing screens, we're designing systems (to be honest this is not new, or it really should NOT be to anyone worth their salt in UX, but there has been a deepening if you will)
We all have our minds on:

  • What should the system notice?

  • What should it ignore?

  • When should it act?

  • When should it ask a follow-up?

  • When should it say, “I don’t know”?

  • When should a human step in?

You can’t solve that with a polished Figma file and a Jira ticket.

You need a shared working model of the experience.

Something design, engineering, product, and data can poke at together.

Something you can test against real scenarios, not just admire in an ideal state.

Handoff collapses when the product starts thinking

Traditional software was mostly deterministic.

A user does X.

The system does Y.

Messy, sure. But still relatively containable.

AI-native products are different.

They infer. They summarize. They rank. They recommend. They occasionally make things up with the confidence of a mediocre man in a strategy meeting.

So the design work gets bigger.

The interface is no longer the whole experience.

The model behavior is the experience.

The prompt architecture is part of the experience.

The escalation path is part of the experience.

The confidence signal is part of the experience.

The failure state is very much part of the experience.

And if design only shows up at the end to “QA the UI,” we’ve already missed the point.

What replaced handoff for me is not designers becoming engineers.

It’s shared authorship.

Designers getting closer to the system.

Engineers getting closer to the experience.

Product getting clearer about what “good” actually means.

Fewer baton passes.

More working loops.

Less “here’s the mockup.”

More “let’s test what the thing actually does.”

That’s the shift.

The new artifact isn’t a handoff file.

It’s a learning loop.

Intent → behavior → user response → evaluation → adjustment.

Again and again.

This is where design leadership is going.

Not toward less craft.

Toward deeper craft.

Less decoration at the end.

More judgment at the center.

Handoff isn’t dying because design matters less.

It’s dying because the wall between design and engineering is finally too expensive to maintain.

And honestly?

Good.

Read more

Read more

Yvonne Doll

UX, Product Design, AI Systems

Share post

More

More

And what comes next. Design/Engineering Co-authorship

And what comes next. Design/Engineering Co-authorship

Demos Are Theater Demos are seductive because they make everything look easy. But let’s be honest: a demo is a theater.

Demos Are Theater Demos are seductive because they make everything look easy. But let’s be honest: a demo is a theater.

You've heard it, I've heard it, heck, we've probably all said it (and really meant it.) "Let's ship and learn". Every org claims to want to "ship and learn." The problem is: shipping is easy to measure (did it go live?) While learning is not. Teams celebrate velocity while skipping the critical feedback loop that tells you if what you shipped worked, and what to do next.

You've heard it, I've heard it, heck, we've probably all said it (and really meant it.) "Let's ship and learn". Every org claims to want to "ship and learn." The problem is: shipping is easy to measure (did it go live?) While learning is not. Teams celebrate velocity while skipping the critical feedback loop that tells you if what you shipped worked, and what to do next.

AI isn’t here to replace designers, it’s here to clear the clutter.

AI isn’t here to replace designers, it’s here to clear the clutter.