AI Categorized 2,400 Expenses Last Quarter - Zero Manual Sorting (8 Nodes) 🔥

Finance team categorizing expenses manually. Every receipt. Every transaction. Someone reads merchant name. Decides category. Types it in.

Built 8-node workflow. 2,400 expenses last quarter. Zero manual sorting. AI handles categorization.

THE CATEGORIZATION NIGHTMARE:

Is Uber travel or transportation? Is that Amazon purchase office supplies or equipment? Starbucks - meals or client entertainment?

Every person categorizes differently. Audit comes. Categories inconsistent. Explanation needed for every decision.

THE DISCOVERY:

Document extraction returns category as enum. AI picks from predefined list. Same logic every time.

But AI isn't perfect. Added merchant keyword fallback. Uber always maps to Travel. Starbucks always maps to Meals. Consistency wins.

THE WORKFLOW:

Google Drive trigger watches receipts folder → Download expense document → Document extraction pulls merchant, amount, date, category → Code processes and applies keyword fallback → Sheets logs to expense tracker → IF checks threshold → Over $500 triggers Slack approval → Under $500 confirms logged.

8 nodes. Categorization automated.

THE SMART CATEGORIZATION:

JSON Schema includes category enum: Travel, Meals, Office Supplies, Software, Equipment, Marketing, Professional Services, Utilities, Other.

AI picks best match. Code validates against merchant keywords. Known merchants get consistent categories regardless of AI interpretation.

THE APPROVAL ROUTING:

Threshold-based logic. Anything over $500 requires manager approval. Slack notification with full details. Everything else auto-logged.

No more $3,000 "office supplies" purchases slipping through. Manager sees it immediately.

THE TRANSFORMATION:

Before: 15 minutes per expense report reviewing categories. Arguments about what counts as what. Inconsistent across team.

After: Zero time categorizing. Consistent logic. Manager only reviews high-value items.

THE NUMBERS:

2,400 expenses processed last quarter