How to extract invoice data into a bookkeeping sheet with n8n and OpenAI

Finance & invoicingn8n

The problem

Receipts and invoices arrive as photos and PDFs. Typing date, merchant, amount and VAT into a spreadsheet and filing the file is slow and error-prone.

Tools needed

• n8n
• OpenAI (vision)
• Google Drive
• Google Sheets
• A simple upload web page

Step-by-step recipe

1. Build a small upload page that posts the file to an n8n webhook.
2. Identify the user (the prototype looks up a token in a Google Sheet).
3. Send images to OpenAI vision; for PDFs, extract the text first and then analyse it.
4. Have the model return JSON (date, merchant, gross amount, VAT rate, category) and show it to the user to review and edit.
5. On save, store the file in a Drive folder per user and month, and append the data to a yearly sheet.

Copyable prompt

Extract bookkeeping data from this invoice or receipt. Return JSON only:
{"date": "YYYY-MM-DD", "merchant": "...", "gross_amount": number, "currency": "...", "vat_rate": number|null, "vat_amount": number|null, "category": one of [CATEGORIES], "invoice_number": "..."|null}
Use null for anything not clearly printed on the document. Do not calculate or guess missing values.

What the builder reported

A working prototype with a demo video. The source gives no accuracy or time-saving figures, and the public repo contains sanitised exports that need your own credentials and IDs.

These are the source author's own claims. AgentRecipes has not tested or verified them.

Watch out for

• The token lookup is a stand-in for real authentication; don't expose it publicly as is.
• Keep the human review step: extracted amounts and VAT rates must be checked before they reach your books.

Source

More finance & invoicing recipes