I started forgetting autolayout

Notes on the AI-rearranged workflow, from yours truly.

A few months ago I noticed I opened Figma less all week and I started forgetting the joy of autolayout. 

I've been in Claude, Cursor, Gong, and Perplexity, not necessarily in that order, and somewhere between the second and the third I'd talked to a real customer. The work was getting done. It just wasn't being done the way I'd learned to describe it in interviews.

Workflows change every day so what I’m writing now might not age well, but I think it’s in good company. 

What changes isn't the work, it's the medium I get to think in. For a decade, designers explored in static tools (Sketch, then Figma) and shipped work that engineers had to translate into the actual product. That was the handoff economy. Had its issues, but was okish.

AI didn't kill that arrangement, but it reshaped it. Now I can prototype in the medium the design will be rendered in. I can talk to a customer on Tuesday, prototype the answer on Wednesday, and have something to put in front of them by Friday (unless priorities change).

The new order of operations

I start in Claude. Not for the design, for the question. I write a short brief, name the edge cases I'm worried about, list the assumptions I want to test, and use the result to align with my PM and engineers in one conversation instead of three. The PRD isn't always good. But it's specific enough to disagree with, which is the only thing a first draft needs to be.

Then I look at what's already there. Gong gets me into customer calls I'd never have time to sit through and its summaries surface patterns across thirty conversations in the time it would have taken to listen to two at 1.5x. Perplexity does the breadth work: where's the landscape sitting, where am I about to reinvent something that already exists.

I talk to customers directly. The point isn't to validate but to change my mind quickly when I'm wrong. Research should move the work forward, not freeze it. The point isn't to earn permission to design, it's to find out where I'm wrong fast enough that it doesn't cost weeks.

Then I prototype. Cursor turns the flow into working code I can actually click through. The output isn't production-ready and doesn't need to be. It needs to be touchable enough that a customer can poke at it and tell me what's confusing, something a Figma prototype, however polished, can't quite do.

I polish the UI and document the edge cases. And whatever shape the workflow took, I write up what worked for the team: prompts, patterns, edge cases I hit, so the next person doesn't start from zero.In an age where anyone could be a one-person band, I truly believe collaboration matters most.Sharing with your team should be a must. There is no reason for gatekeeping.

None of this is a stack. The tools above will be different in a year (more likely a week), and they'll be different again at the next company. They're a set of choices, not a workflow.

Where the decisions go

Not all decisions belong in the same place. Some go to a tool. Some go to a teammate. Some stay with me.

• Range and breadth. What's the landscape, what's already been tried, where am I about to reinvent something. I let an LLM cover this. Perplexity gives me the survey, Claude helps me structure what I find.

• Execution and repetition. Scaffolding a flow, writing the boilerplate, getting a prototype touchable. Cursor handles the code, I shape the behaviour and decide what's worth keeping.

• Judgment. Which gap matters to my user. What to ship, what to cut. That stays with me. Not because nothing else could try, but because if I outsource the decisions that define the work, I'm not really the designer of it anymore.

The tools will shift; that map mostly won't. The skill is knowing which row a decision sits in before reaching for the LLM. (Or, you know, taste, if you want to call it that.)

What hasn't changed

The risk in any of this is becoming a middleman: someone whose week is mostly moving outputs from one tool to another, approving as you go. If that's all I'm doing, I'm packaging, not designing. The hard problems, the ones that build the muscle I'll need for next time, never actually pass through me.

The other thing AI hasn't changed: the people I'm designing for. They still don't care whether a prototype was hand-coded or hand-drawn. They care whether it survives a Tuesday 7pm shift, when something's already gone wrong and the queue isn't going to wait while I figure it out. That's still the test.

What's still missing

The piece of this workflow I haven't looked at yet is AI for Handoff documentation. 

We still need it, maybe more, not less, if we're moving faster. 

As much as I involve my devs since the very beginning of a project and make sure to know and understand the constraints I still think that documenting decisions and reasons makes a big difference. Teams that ship well don't keep things in their heads or in a small circle; they write them down, audit them, share them. 

To Conclude

There is so much noise on social media about how AI is killing design, a lot of clickbait and rage bait articles which are rooted on how as humans we’re drawn to anything that we “fear”. I’m not delulu, however I can choose how to look at the future and hope to find like-minded individuals.

None of this changes what the role is. I'm still a product designer (or whatever cognitive shortcut that makes us feel safe and in control we wanna call it - maybe a debate for another day y’all). I still spend most of my time thinking about people who haven't been given the experience they deserve. What's changed is what I can do in a week and where I can afford to spend the time I save.