It has been a fun few weeks getting to know some of the latest AI tools through my work on the puzzle counter app (Part 1, Part 2). I will use this post to venture some guesses about how this will affect product development, with a specific focus on how it will change design. It was eye-opening to see how quickly the tools at our disposal have evolved from beguilingly smart tchotchkes into meaningful collaborators for the complex task of product design.
This post promises to be a shorter one. I realized pretty quickly into this stage that my assumptions about how this process would work end-to-end were not correct. I imagined that once I had some sets of wireframes from the initial explorations, I would import those into a tool and voila they would import and provide a the tools necessary to create, edit, and apply a cohesive design system. In reality, this didn’t seem to be the way most tools worked, likely for the best, since there are more straightforward workflows available.
For this first step I wanted to try a few different tools to begin to understand the tools that are out there and have a few data points to compare the output of each. For wireframing, I used two general LLMs (Claude and Chat GPT) and one design-specific tool (UX Pilot). For Claude and ChatGPT I used a method of creating simplified and efficient ASCII wireframes.... For my specific goals, I was especially interested in the performance of the general LLMs since they were obviously not built with design in mind.
