AI Invoice Receipt Extraction from Gmail to Google Sheets (n8n)

AI-powered invoice receipt extraction from Gmail straight into Google Sheets (with human review when needed)

This n8n workflow automatically scans your Gmail for invoice PDF attachments, uses AI to extract key fields (vendor, invoice number, dates, currency, net, tax, total), and logs results into Google Sheets—deduplicated and validated. If anything looks doubtful, it emails you for review as “Needs review”, keeping your bookkeeping accurate and audit-ready.

What this workflow does

• Polls Gmail for the right emails: Only messages with real PDF attachments pass through—newsletters and calendar invites are guarded out so they don’t trigger AI.
• Extracts PDF text inside n8n: The workflow extracts the PDF’s text directly within n8n (no OCR service and no per-page OCR fees).
• Analyzes invoices with AI: The AI returns strict JSON including vendor, invoice number, invoice/document type, dates, currency, net, tax, total, and a confidence score.
• Validates numbers in code: A Code node normalizes currency and checks arithmetic by validating net + tax vs total.
• Deduplicates automatically: The workflow looks up the invoice number in your sheet before writing—forwarded copies or reminder emails won’t be logged twice.
• Flags uncertain results: Low-confidence or broken extractions (e.g., missing total or failed validation) are emailed to you and logged as Status = “Needs review”; clean records are logged as Status = “Auto”.

Use cases

• Bookkeeping teams who want invoice receipt automation without manual data entry.
• SaaS operators consolidating invoice data from forwarded Gmail messages into a single Google Sheets ledger.
• n8n engineers building reliable AI invoice extraction with code-based validation and an exception workflow.

Technical details

• n8n nodes: if, code, no op, Gmail nodes, plus supporting nodes like sticky note (for clarity).
• Inputs: Gmail via “When Invoice Email Received” and a review alert to your address.
• AI: AI Invoice Analyzer using an OpenAI (or compatible) credential.
• Output: One Google Sheets table with columns such as Date, Vendor, Invoice number, Invoice date, Type, Currency, Net, Tax, Total, Status, and Email subject.