Invoice PDF to Excel Extraction with GPT-4o-mini in n8n

Automatically extract invoice data from PDF to Excel in n8n using GPT-4o-mini

This n8n workflow scans a local folder of invoice PDF files, extracts structured header + line-item data using OpenAI (gpt-4o-mini), validates required fields and subtotal math, and outputs per-invoice Excel files—while archiving PDFs into valid, invalid, or error folders and generating a daily execution report.

What this workflow does

• Runs on a daily schedule and checks that n8n is self-hosted with filesystem access to your configured local directories.
• Creates an initial execution-report Excel file to track totals.
• Reads each invoice PDF from the configured local input folder and extracts text from the document.
• Handles unreadable PDFs by routing them to the error workflow path.
• Cleans extracted text and sends it to OpenAI (gpt-4o-mini) with a strict JSON schema to extract invoice header fields and line items.
• Normalizes the AI response, then validates required fields and checks line-item/subtotal consistency.
• Formats and writes results into an Excel output per invoice in the local output folder.
• Routes each source PDF into valid, invalid, or error folders, and removes it from the input folder.
• Overwrites the execution report with final counts and end time.

Use cases

• Converting scanned or formatted invoice PDFs into consistent Excel spreadsheets for bookkeeping.
• Batch processing daily invoice intake with automated validation to catch missing fields or incorrect totals.
• Maintaining an auditable workflow by archiving PDFs based on extraction and validation results.

Technical details

• n8n Community Edition self-hosted (local file access required)
• Built with workflow nodes including if, set, code, merge, and switch (plus sticky note for clarity)
• OpenAI (gpt-4o-mini) for structured JSON extraction with schema constraints