Jon Cheney: Product design in the age of AI

Over the last year I’ve been sort of freaking out as a product designer. If that’s how you've been feeling please read on and checkout this video.

With the advent of AI tools, chatbots, agents, and new releases of tools every month, I’ve been trying to wrap my head around how to benefit from potential improvements to workflows, and not be swept away by the shifting tech landscape. I've also been thinking about others, designers who are more junior, and what work can look like for them going forward.

At first I was in denial, thinking it was all a bit silly, that this would merely be the latest in a long line of fads. The experiences I’d honed over a lifetime in design could surely beat anything done by AI.

But the incremental changes released by these massive LLMs started to get better and better.

Bit by bit small pieces of my work became replaceable. Has that happened entirely? No, not yet. But more and more parts of the work of design can be swapped out by a well written prompt. 

So I set out on a path this year to try and understand how I can benefit from AI tools, and where I can use them to build better products. 

Some of this, frankly, feels like a hype game. Not too long ago we were all sold on the metaverse, cryptocurrency, heck; I’ve been around long enough to have bought into the Beanie Baby craze.

If you’re a UX or product designer, website designer, or designer of basically any type, you can still find work; there’s still things to do out there. The world is a huge place and people of all types need design at different capacities. 

But you’re surely sensing it. A shift in how people think about the services they’d typically pull designers in for, turning to chatbots and AI tools to figure it out themselves. 

So a few weeks ago, when I was offered the chance by leadership at Angel to take a course on generative AI, I jumped at the opportunity. Well, sort of. I walked in with some skepticism, but also knew that this is the type of thing I have to figure out, even if I think parts of it are overhyped. 

The course started out with things I already knew as as the basics (explaining the various chatbots and how you use them), and at first I thought that’d be it. But then Jon Cheney walked through the steps to using AI to prompt yourself. That was something I hadn’t been trying. When you have an idea, ask ChatGPT (or your favorite chatbot) to interview you, one questions at a time, until it has enough information to generate a good prompt; then toss that back in and see what happens. That simple process shifted how I have approached every interaction with chatbots since. 

And I haven’t been able to stop writing about it. 

A friend of mine just asked for advice on how to build a quick website for an upcoming conference. I walked him through the basic prompt loop that Jon describes, and then dropped the results into Grok, Claude, and ChatGPT. Between the three of them we got some pretty decent design ideas back, which ultimately led into my suggesting he find a Framer template and modify things for himself.¹

If you’re at all nervous about where design is headed, and trying to figure it out for yourself, then I recommend giving this process a try (and watching the video between me and Jon).

Nobody knows the future, and you’re not along in your concerns and uncertainty. But for now, at least in this moment, I’ve found something to focus on, to try, to be excited for. And frankly, I haven’t been able to stop tweaking and running through prompt ideas, testing, and seeing what spits out on the other side.  

And if you're like me and prefer podcasts over YouTube, checkout the RSS feed and download on your favorite podcast player.

Join the paid newsletter and get 3-4 extra posts like this each week!

Subscribe Now

1. And this is where AI isn’t perfect. Some things it struggles with. If you want to build a basic portfolio site it’s perfect to prompt yourself, run through the interview process, and see what spits out. You can use that output to test in the different chatbots, and even throw it into Replit. But, at least mid 2025, it’s still faster to find something like a Framer template and modify the pieces you want to keep.