Work
Complex, high-stakes software, and the way it gets built.

Building belongs inside discovery, so everyone has to be in the room before anything is built
Working with AI gave us time back. Here’s how we’re using it.

From a pile of quote requests to one list of people and homes
When Obie opened its platform to independent agents, the page they’d live in was built for our own brokers. I redesigned what it was organized around. Nearly three years later, it’s one of the most-used pages in the product.

Homeowners ran on a platform built for landlords, so every seam was ours to hide
Obie’s first homeowners line, on a backend that had never contemplated it, and the seams we didn’t hide until agents found them.

The quote is the product, so the quote belongs on the page
On building an insurance quote like a Ferrari.


When the core can’t change, design owns the path and the record
Expert users, an immovable third-party core, and a design that survived losing its favorite idea.


It looks like a prototype. It’s built like a product.
How I built the system behind design’s working prototype: a public queue for design work, a written record of every decision, and a rulebook that governs AI sessions.

The behavior is the product, so the prototype has to run
In insurance software, what a screen does is usually a rule. A frame shows what the screen looks like. It leaves the rule out.
The diff that audited the design system
Machine-checking design tokens found a bug every human eye had passed for years.
Delegate the labor. Write down the judgment.
Agents made labor cheap. Judgment costs what it always did, and it doesn’t delegate.