The context
VendorOS exists to solve a scale problem: a US government agency needed to manage the complete lifecycle of its vendors and contractors — onboarding, active management, offboarding — across five distinct roles within the organization, each with its own permissions, workflows, and information needs. Before this project, that process lived scattered across spreadsheets, email, and manual document review.
My role
I led end-to-end product design: information architecture for the five roles, the critical flows for each stage of the lifecycle, and the component system that holds this much information density without becoming overwhelming.
Trustworthy AI, not AI for its own sake
The most delicate part of the project was AI-assisted extraction of data from vendor documentation. The design premise was simple but non-negotiable: the AI suggests, a person decides. Every extracted field shows its source evidence — which document it came from, at what confidence level — and nothing is saved as final until someone reviews and confirms it. In a government context, traceability isn’t a UX nicety; it’s a requirement.
The outcome
A single platform replaced a process scattered across multiple tools and manual reviews, with faster onboarding thanks to AI-assisted extraction — without losing human control at any step where it matters.