Extract Invoice Data in n8n Without Uploading Files to Third Parties

The Problem: Invoice Data Entry in Automated Workflows

You have an n8n workflow that processes orders, syncs accounting data, or manages AP. But invoice PDFs still need manual data entry because:

• Cloud OCR services require uploading invoices — a privacy and compliance concern for financial documents
• Tesseract alone doesn’t understand invoices — it reads characters, not structured fields like line items, totals, or tax rates
• Per-page pricing adds up fast — Rossum, Nanonets, and Mindee charge $0.05–$0.50 per page, which compounds on batch runs
• Self-hosted LLM extraction is slow and unreliable — running Ollama for document understanding needs GPU and still hallucinates amounts

n8n Workflow: Invoice Email → Extraction → Google Sheets

Here’s the practical setup. This workflow watches an inbox for invoice PDFs, extracts the data, validates the math, and appends a row to Google Sheets.

1. IMAP Email Trigger

Configure the IMAP node to watch your AP inbox. Filter for emails with PDF attachments — most invoice emails have the PDF inline or attached.

{
  "node": "IMAP Email",
  "parameters": {
    "mailbox": "INBOX",
    "options": {
      "allowUnauthorizedCerts": false
    }
  },
  "note": "Filter downstream for .pdf attachments"
}

2. HTTP Request — Extract Invoice

Send the PDF binary to the extraction API endpoint. The response is structured JSON you can immediately use in downstream nodes.

{
  "node": "HTTP Request",
  "parameters": {
    "method": "POST",
    "url": "https://api.useful-patch.com/v1/extract",
    "sendBody": true,
    "bodyParameters": {
      "parameters": [
        {
          "name": "file",
          "parameterType": "formBinaryData",
          "inputDataFieldName": "attachment_0"
        }
      ]
    },
    "headerParameters": {
      "parameters": [
        {
          "name": "Authorization",
          "value": "Bearer YOUR_API_KEY"
        }
      ]
    }
  }
}

3. Example API Response

{
  "vendor": "Acme Supplies Ltd",
  "invoice_number": "INV-2026-0847",
  "invoice_date": "2026-04-10",
  "due_date": "2026-05-10",
  "currency": "GBP",
  "subtotal": 840.00,
  "tax_rate": 20,
  "tax_amount": 168.00,
  "total": 1008.00,
  "line_items": [
    {
      "description": "Office chairs (ergonomic)",
      "quantity": 4,
      "unit_price": 160.00,
      "line_total": 640.00
    },
    {
      "description": "Standing desk converter",
      "quantity": 2,
      "unit_price": 100.00,
      "line_total": 200.00
    }
  ],
  "confidence": 0.97,
  "validation": {
    "line_items_sum_matches_subtotal": true,
    "tax_calculation_matches": true,
    "total_matches": true
  }
}

4. IF Node — Math Validation

Check the validation object. If all three checks pass, route to the “auto-approve” branch. If any fail, route to a Slack notification or email alert for human review.

{
  "node": "IF",
  "parameters": {
    "conditions": {
      "boolean": [
        {
          "value1": "={{ $json.validation.total_matches }}",
          "value2": true
        }
      ]
    }
  }
}

5. Google Sheets — Append Row

Map the extracted fields to your spreadsheet columns. Each invoice becomes a row with all the structured data ready for reconciliation.

{
  "node": "Google Sheets",
  "parameters": {
    "operation": "appendOrUpdate",
    "documentId": "YOUR_SHEET_ID",
    "sheetName": "Invoices",
    "columns": {
      "mappingMode": "defineBelow",
      "value": {
        "Vendor": "={{ $json.vendor }}",
        "Invoice #": "={{ $json.invoice_number }}",
        "Date": "={{ $json.invoice_date }}",
        "Due": "={{ $json.due_date }}",
        "Subtotal": "={{ $json.subtotal }}",
        "Tax": "={{ $json.tax_amount }}",
        "Total": "={{ $json.total }}",
        "Line Items": "={{ $json.line_items.length }}",
        "Confidence": "={{ $json.confidence }}"
      }
    }
  }
}

Privacy-First Extraction: Why It Matters for Invoices

Invoices contain some of the most sensitive business data: vendor relationships, pricing agreements, payment terms, bank details, tax IDs. Uploading them to a cloud OCR service means:

• The vendor’s data processing agreement now covers your financial documents
• Your supplier relationships and pricing are in someone else’s training pipeline (unless explicitly excluded)
• Compliance teams need to audit every third-party processor that touches financial data

Client-side extraction avoids all of this. The PDF is processed in your browser or on your own infrastructure. Only the structured JSON result — which contains no raw document images — moves through your n8n workflow.

Common n8n Invoice Automation Patterns

Pattern 1: Email Inbox → Extract → Accounting Software

The most common setup. IMAP trigger watches your AP inbox. Invoice PDFs are extracted and pushed directly to Xero, QuickBooks, or FreshBooks via their respective n8n nodes. Validation checks catch OCR errors before they hit your books.

Pattern 2: Watched Folder → Batch Extract → Google Sheets

For teams that collect invoices in a shared drive folder. A cron-triggered workflow scans for new PDFs, extracts each one, and appends rows to a master spreadsheet. Good for monthly reconciliation or audit prep.

Pattern 3: Webhook → Extract → Approve → Pay

Your procurement system sends a webhook when a new invoice is received. n8n extracts the data, checks it against the purchase order (amount match, vendor match), routes for approval via Slack or email, and triggers payment on approval. Full AP automation loop.

Pattern 4: Multi-Format Intake → Normalize → ERP

Invoices arrive in different formats from different vendors — some PDF, some email body, some scanned images. n8n routes each type through the appropriate extraction path, normalizes the output to a common schema, and pushes to your ERP. The extraction API handles layout variation automatically.