Automate invoice review and purchase orders in n8n

An n8n flow checked invoices against stock records, then filed purchase orders and alerts for a client.

Done with n8n· Intermediate

What they did
A 23-node n8n setup kept an eye on the client’s Gmail for invoice messages. When one arrived, it pulled the PDF, fed it to OpenAI’s vision model, and captured the vendor, total, line items, and dates. The workflow then compared those details with a Google Sheets inventory list; if the match looked good and the totals seemed plausible, it generated an ERP purchase order, sent a Slack update, and replied to the sender.
How it went
The client shifted from handling 40-50 invoices a week by hand to roughly 90% automatic processing. That freed up about 8-10 hours each week. The builder said the work took 3 days, with most of the time spent on edge cases, and the handoff to the operations manager took around 2 hours.
Tips
Expect edge cases to take up most of the build. Also watch n8n usage closely: the builder said each invoice burns about 23 executions, so at 200 invoices per month the Pro plan is the lowest viable tier.

Try it yourself with n8n

Create an n8n workflow that watches [email inbox or folder] for incoming invoices, extracts the vendor, line items, quantities, amount, and date from each PDF, and checks those fields against [Google Sheets inventory link]. If the invoice matches inventory and the figures pass a reasonableness check, prepare a purchase order in [ERP or purchasing tool] and draft a notice for [Slack channel or email contact]. Before any payment, send, submit, delete, or other irreversible action happens, pause and show me the exact order and message for approval. Put anything uncertain into [review list or person] instead of acting on it automatically.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on dev.to ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.