Order Management System
A RevOps app that reads sales notes with AI and reports per-deal economics.
Problem
Custom orders moved through engineering, design, and clients on scattered spreadsheets and long email threads. Nobody could see per-deal economics, and approvals dragged for over a week.
Action
Built a modular Apps Script backend with a sidebar UI. Schema-constrained LLM extraction (Claude and Gemini) turns raw rep notes and email threads into structured order data. A RevOps dashboard reports per-deal gross profit, status buckets, and an approval-speed KPI computed at the deal level — deliberately excluding prospects and lost deals, because a lost deal's approval time isn't a real outcome. Also produced a documented migration-readiness assessment mapping the spreadsheet model to Postgres tables and flagging normalization gaps.
Impact
- project approval time cut from 8 days to 5 or fewer
- ~60% less manual work
- 40+ upsells closed, ~80% success rate
- $15K+ incremental gross profit across three brands