CASE STUDY | SHORT READ
Line item extraction: AI powered document review experience
Replaced paid converters and manual data entry with an AI-powered review flow that accountants actually trust — saving ~50,000 accountant hours every month inside QuickBooks.
My Role
Team
Timeline
Platform
IMPACT
THE 30‑SECOND VERSION
THE TURNING POINT
Accountants already had tools that hit near‑100% accuracy. In interviews, the same thing kept surfacing, they weren't struggling to read the extracted data. They were deciding whether to believe it.
WHAT THAT PRODUCED
Trust‑first isn't a slogan — here's what it looks like as a screen.
THE CALLS BEHIND IT
Four decisions did most of the work — spanning strategy, interaction, systems, and scope.
Each decision's full option space — plus the selection model, date‑picker constraint, and change‑view toggle — is in the full case study.
HOW I DE-RISKED IT
None of those calls were hunches. Three focused interaction studies — on top of interviews and a competitive teardown — settled them before a line of production code.
WHERE IT LANDED
Back to the number we opened on — here's what it added up to.
Each study produced a decision for this project — but the underlying reasoning generalizes. These are the adaptations worth applying to the next dense, verification‑heavy interface.
