For the past year, I have not opened Figma. I plan features and build interactive, production-shaped prototypes directly in code with AI — going from a rough idea to something a support advocate can actually click through in the same sitting a static mock would have taken to draft. Kraken, Big Otto, and Otto Aha (shown below) are a few of the dozens of prototypes built this way.
client
Gusto
year
'26
timeframe
Ongoing
Problem
A picture of an interface is a proxy for the experience, not the experience itself. Stakeholders react to what they imagine an interaction feels like, and the friction that actually matters — loading states, edge cases, information density under real data — only shows up once something is running. By the time that friction surfaces in a review of static frames, a design has often already been approved and staffed.
Working this way means user testing happens the same week as the idea instead of the sprint after handoff, and the feedback that comes back is about the real thing instead of someone’s best guess at how it might feel.
Solution
I build working prototypes directly in code with AI coding tools, reusing the same component patterns and interaction logic the eventual product will ship with. Kraken, Big Otto, and Otto Aha (shown above) are a few of dozens of prototypes tested with advocates before a single line of production code was written.
I am also responsible for merging design PRs directly — giving me full visual control, elevating the design bar, and increasing team productivity. Our SDLC is now dramatically shorter thanks to this tighter design/engineering partnership.
Kraken — Case Routing Panel prototype
Big Otto — full Salesforce sandbox with Otto panel
Otto Aha — contextual chat highlight and instructions
PD
AI
ENTERPRISE