Design Agent · JLL

Corrigo Prototype

Corrigo Enterprise was designed in Figma and built in a codebase that runs legacy pages next to newer React ones. A Figma file shows one state at a time, drifts from what is live, and keeps none of the reasons. For the Dispatch Board, Schedule Agent, and the Work Order List, I moved the design out of Figma into working prototypes in code, built with an AI coding agent, close enough to production to review against it.

Company
JLL
Product
Corrigo
Role
Design Lead
Period
2026

Decisions

01

Build it the way production builds it.

Legacy pages use the same stack and page setup as production, the Dispatch Board is React like its rewrite, and every route follows production's path. Any screen links straight to the one it mirrors, and engineers read code they recognize.

02

The live product wins.

When sources disagree, the QA environment comes first, then the latest acceptance criteria on the ticket, then Figma. I measure colors and spacing from the live page instead of matching by eye, because an eyeballed match once picked white where the product uses a grey.

03

Options sit side by side in the build.

A strip above the app switches permissions, admin settings, and design options live, so reviewers compare one option with another on the real screen instead of on two frames. Work Zone Priority went from two options to transfer lists, then lost drag and drop, because real lists run to hundreds of work zones and the developers asked for it.

04

Every decision is numbered and kept.

A change gets a new decision that supersedes the old one, and a missing reason is written down as missing instead of guessed. The before states here were rebuilt from the old commits, so they are real renders.

05

The agent stays on the design system.

Rules files the agent loads require design system components and tokens and no hard-coded colors. It checks the running prototype in the browser at real widths before I call a change done.

Outcome

I led the Corrigo Enterprise prototypes as Design Lead, with a second designer contributing, and built most changes with Claude Code in small pull requests. Reviewers and engineers work from a running prototype that matches the live product, with the reason for each choice in the repo. It is the reference for the Schedule Agent and Dispatch Board work engineering builds next.

Screens

Schedule Agent's constraint profiles and the profile editor, Work Zone Priority from its two early options to the transfer lists it settled on, and the AI assistant page in the back office.

The screens are behind a password

The screens for this and the other case studies are shared with the people I am talking to. The password is in the resume I sent with my application. If you do not have it, ask me for it.