Tags: Enterprise AI-Assisted Complex Systems ServiceNow Sole Designer
Client — Genpact (Internal Product) Product Type — AI-assisted commercial underwriting platform My Role — Sole UX designer, end-to-end Platform — ServiceNow Timeline — 8–9 months (ongoing) Modules — Exposure Management · Risk Assessment · Quote & Bind
Screens shown are selectively cropped to protect proprietary product design. Full work available on request.
Commercial insurance underwriting is one of the most document-heavy, decision-dense workflows in financial services.
A single submission involves dozens of documents — schedules of values, loss reports, inspection records, sanctions checks — arriving incrementally over days or weeks. Before any risk can be assessed or policy priced, all of it has to be read, extracted, validated, and structured.
The platform would use AI to handle the heavy lifting. Three modules. Designed from scratch. In a domain that was completely new. On a platform with real constraints.
Four problems defined the entire design challenge.
Problem 1 — AI that nobody trusted The system extracted data automatically but showed no reasoning. Underwriters either trusted blindly or re-read every document manually. Neither was acceptable at scale.
Problem 2 — Risk signals buried behind clicks Critical information existed across every module — but wasn't surfaced. Sanctions hits summarised. Guideline failures undifferentiated. Authority breaches implicit. Reviewers were finding out about risk after navigating past it.
Problem 3 — No map, no blueprint, no prior version There was no existing flow to build on. No competitor to reference. Requirements arrived incomplete and kept changing. The entire platform had to be designed from scratch in an unfamiliar domain, on a constrained platform, while the brief was still being written.