Optimize Invoices: PDF to JSON with n8n OCR Automation

Transform Invoice Processing with Intelligent OCR-to-JSON Automation

Stop manually extracting data from invoices and documents. This powerful n8n workflow combines Mistral AI's OCR capabilities with intelligent LLM processing to automatically convert your PDF invoices and document images into clean, structured JSON data—complete with confidence scoring and quality control.

What This Workflow Does

This OCR automation workflow streamlines document processing through six precise steps:

• Accepts PDF or image uploads via webhook as binary data
• Processes documents using the Mistral OCR node for text extraction
• Normalizes and cleans the raw OCR output
• Sends processed text to Mistral AI's LLM for structured JSON extraction
• Applies final cleaning and normalization to ensure data quality
• Returns intelligent status codes: "ok" for high-confidence results or "review_needed" for manual verification

Use Cases

Perfect for automation engineers and SaaS operators who need reliable document processing:

• Invoice automation: Extract vendor details, amounts, dates, and line items from supplier invoices
• Receipt processing: Digitize expense receipts for accounting systems
• Purchase order management: Convert PO documents into structured data for procurement workflows
• Form digitization: Process scanned forms and applications into actionable data

Technical Details

Built with proven n8n nodes for enterprise reliability:

• Webhook node: Secure file upload endpoint at /webhook/ocr-to-json
• Mistral AI integration: Advanced OCR and LLM processing with configurable confidence thresholds
• Code nodes: Custom normalization and data cleaning logic
• Conditional logic: Intelligent routing based on processing confidence scores

Setup is straightforward—import the workflow JSON, configure your Mistral AI credentials, activate the workflow, and start processing documents immediately. The workflow includes built-in quality control, automatically flagging low-confidence results (below 0.5) for human review, ensuring accuracy in your automated document processing pipeline.