n8n + AI Vision — Automated Invoice Processing (PDF → ERP) Supplier invoices arrive as PDFs in a shared inbox. AI Vision reads each invoice, extracts header, line items and totals, validates the math, and posts the entry directly to the ERP — SAP, NetSuite, Dynamics, Odoo or Sage. Exceptions only escalate when truly needed. The problem Accounts payable was bottlenecked by inbox triage. Someone had to open every supplier invoice PDF, retype headers, line items, VAT and totals into the ERP, and chase the supplier when numbers didn't match. Slow, error-prone, and the backlog built up every month-end.

What I built A fully automated invoice-to-ERP pipeline that handles the full document flow without a human touch unless something is genuinely off:

Inbox watcher detects new invoices in the AP shared mailbox.

AI Vision (GPT-4) parses the PDF — supplier data, invoice number, dates, line items, subtotals, VAT, totals.

Validation layer checks the math, matches the supplier against the master vendor list, flags duplicates and missing fields.

ERP posting — entry written directly into SAP / NetSuite / Dynamics / Odoo / Sage via official API or DB connector.

Exception queue — only documents that fail validation hit a human, with the reason pre-explained.

The stack n8n · GPT-4 Vision (OCR + structured extraction) · ERP connectors (SAP / NetSuite / Microsoft Dynamics / Odoo / Sage) · Validation engine · Slack/email exception alerts · Audit-ready logs.

Outcome

+400 invoices/month processed end-to-end without human entry

0 typing errors on posted entries — the AI doesn't fat-finger numbers

Month-end closes faster — the AP backlog stops being a bottleneck

Same-day visibility on flagged exceptions Want something similar? The same engine handles purchase orders, delivery notes, contracts, customer onboarding forms — anywhere a structured document arrives by email and gets retyped into a system today.