Flag duplicate and risky AP invoices with Gmail, OpenAI and Supabase

🚀 How it works

Monitors your AP inbox for incoming invoices, extracts structured data with AI, runs duplicate and vendor history checks against Supabase, then scores each invoice for fraud risk — routing suspicious ones to Slack and your AP team before any payment is processed.

- 📬 Gmail Trigger monitors your accounts payable inbox in real time - 🤖 AI Agent extracts invoice number, vendor, amount, currency, dates and line items into structured JSON — no manual data entry - 🔍 Checks Supabase for duplicate invoice numbers already in the system - 🏢 Checks vendor payment history — flags unknown vendors and amount deviations above 50% from the vendor's historical average - 🧠 Second AI Agent scores fraud risk: low / medium / high / critical with specific fraud indicators and a recommended action - 🚨 High/critical risk — posts a detailed Slack alert to invoice-alerts and emails your AP manager with a hold notice - 🗄️ Logs every processed invoice to Supabase with risk score and status

🛠️ Set up steps

Estimated setup time: ~20 minutes

1. Gmail Trigger — connect Gmail OAuth2; point it at your AP inbox 2. OpenAI — connect OpenAI API credential (used by both AI Agent nodes) 3. Supabase — connect Supabase API credential; create two tables: invoices (invoicenumber, vendorname, amount, status, risklevel, createdat) and vendors (vendorname, avgamount, totalinvoices, flagged) 4. Slack — connect Slack OAuth2; update the channel name invoice-alerts 5. Gmail (Send) — connect Gmail OAuth2; replace ap-manager@example.com 6. Follow the sticky notes inside the workflow for per-node guidance

📋 Prerequisites

- Gmail account receiving invoices - OpenAI API key (GPT-4o) - Supabase project with invoices and vendors tables - Slack workspace with an alerts channel Custom Workflow Request with Personal Dashboard