How I work with AI
It's the question I get asked the most. The short answer: AI expands how much I can explore and iterate. It doesn't replace knowing where we're going, or why one idea is better than another.
I use AI to generate explorations, variations and iterations at a fast pace. Then I often throw 80% away. But what's left is where I start the real work of tweaking, improving and pushing the work.
My job is to set the direction, frame the problem, and decide what's good. I edit, combine and push the work until it's right.
The setup
I connect everything through a CLI, usually Warp. From there, work flows out to design tools for exploration and systems, and to the browser for prototypes and artifacts. Claude is my collaborator across all of it.
Design
Early concepts and high-fidelity exploration.
Design system organization, focused work and production handover to engineering.
Browser
Interactive builds for user testing.
Living documents that communicate ideas with engineering.
A typical project
I frame the problem, set constraints and define what good looks like. The clearer the brief, the better the output.
Claude generates early concepts and high-fidelity explorations. I react, edit and push toward a direction.
The chosen direction becomes components and tokens, organized and prepped for engineering handover.
Prototypes go in front of users, and artifacts align the team with engineering on what we're building.
It's clear that we as designers will never work the same way again. Many assume that there are only two options, resist or dive in head first. But I believe that there is a third, more sustainable way to work: a partnership that blends the speed and volume of AI, with the taste and curation of our human minds.