Automated invoice payment tracking with OCR, Claude AI, Slack & Notion DB
How it Works This workflow automates Invoice & Payment Tracking (with Approvals) across Notion and Slack. - Ingest — You drop invoices/receipts (PDF/IMG/JSON) into the flow. - Extract — OCR + parsing pulls out key fie...
Template notes
How it Works This workflow automates Invoice & Payment Tracking (with Approvals) across Notion and Slack.
- Ingest — You drop invoices/receipts (PDF/IMG/JSON) into the flow. - Extract — OCR + parsing pulls out key fields (invoice no, vendor, currency, totals, receipt paid amount/date). - De-dup & Match — We canonicalize vendor + invoiceno and search Notion: - Primary match: Invoice No (+ optional Currency / Vendor (Canon)). - Fallback: uses document Amount Total and dates.
- Decide the action - createunpaid — new invoice (no payment). - createpaid — new invoice fully paid (unverified). - createpartial — new invoice with a first partial payment. - updatepartial — add a partial to an existing invoice. - updatemarkpaid — mark existing invoice paid in full. - manualreview — currency mismatch / overpayment / ambiguous. - archive — push to archive logs (from manual review).
- Slack approvals (one-click) — A message shows previous paid, this receipt, new total, and Approve buttons (links to a Wait for Webhook resumeUrl). Reviewer picks: Approve Partial / Mark Paid / Manual Review / Archive.
- Notion updates
- We only write editable fields: Paid Amount (number), Status (select), Last Payment Date (date). - Formulas (e.g., Amount Total, Amount Due) recompute automatically. - Receipts are saved in a Receipts DB and related back to the invoice.
- Notifications & duplicates — If duplicates are detected, Slack posts a simple list with clickable invoice names. - Archiving — From Manual Review, Archive goes straight to Archived Invoice DB (and optional Archived Source File DB) as a log entry—no pre-checks needed.
Set up Steps Prerequisites 1. Notion DB 2. 4 Slack Channel (Invoice Input, Notification, Manual Review, Duplicate Alert (Optional)) 3. AI Model (We use Claude 3.5 Haiku, Feel free to use Latest Model) 4. OCR Parsing (We Used ocr.space, Feel Free to Change into any OCR Parsing you have)