Azure Document Intelligence integration

OCR/AI: invoice parsing, receipt scanning, ID extraction, table extraction, custom models

Help CentreConnectors

Overview

Azure AI Document Intelligence (formerly Form Recognizer) is Microsoft's document AI service. The Automize Azure Doc Intelligence connector covers prebuilt models (invoices, receipts, ID documents, layout) plus custom-trained models for org-specific document types.

  • Regions: Azure AI Services are regional. Pick a region near your data. Connector pins to one region per connection.
  • Plans: Free tier: 500 pages/month (S0). Standard tier: pay-per-page based on model.

Setting up the connection

Auth type: API Key.

  1. Azure Portal → Create resource → Azure AI Document Intelligence.
  2. Pick region + pricing tier (F0 free or S0 standard).
  3. Resource → Keys and Endpoint → copy KEY 1 + Endpoint URL.
  4. In Automize, open Settings → Connectors → Azure Doc Intelligence → Add connection. Paste endpoint + API key.

Find your credentials at https://portal.azure.com/.


Rate limits

Free tier 1 TPS. Standard 15 TPS per resource. Long-running operations (custom model training, multi-page analyses) are async.


Data model

Two model classes:

    Prebuilt models – trained by Microsoft on common doc types:
      Read           – plain OCR (text only)
      Layout         – text + tables + structure
      Invoice        – normalised invoice fields
      Receipt        – normalised receipt fields
      ID Document    – passports, driver's licenses
      Business Card  – contact extraction
    Custom models – trained on your labelled samples (Studio UI)

Available operations

8 operations available. Click any row to jump to its detail.

OperationCategoryWhat it does
Analyze Custom ModelGeneralAnalyze with a custom-trained model
Analyze ID DocumentGeneralExtract data from an ID document (passport, driver's license)
Analyze InvoiceGeneralExtract data from an invoice
Analyze LayoutGeneralExtract text, tables, and structure from any document
OCR ReadGeneralSimple OCR – extract text from images/PDFs
Analyze ReceiptGeneralExtract data from a receipt
Get Analysis ResultGeneralGet result of an async analysis operation
List ModelsGeneralList available document models

Operations

Analyze Custom Model

Analyze with a custom-trained model

ParameterRequiredTypeDescription
Model Id model_idYestext
Urlsource urlSourceYestext
Base64Source base64SourceYestext

Tips

  • Use a model trained in Document Intelligence Studio on your labeled samples.
  • model_id: the model's UUID from Studio. Returns fields you defined during training.

Analyze ID Document

Extract data from an ID document (passport, driver's license)

ParameterRequiredTypeDescription
Urlsource urlSourceYestext
Base64Source base64SourceYestext

Tips

  • Normalised: FirstName, LastName, DateOfBirth, DocumentNumber, CountryRegion.
  • Supports passports, US driver's licenses, EU/UK IDs.

Analyze Invoice

Extract data from an invoice

ParameterRequiredTypeDescription
Urlsource urlSourceYestext
Base64Source base64SourceYestext

Tips

  • Normalised fields: InvoiceId, InvoiceDate, VendorName, CustomerName, Items[], TotalTax, AmountDue.
  • Returns confidence per field – set thresholds in your downstream logic.

Analyze Layout

Extract text, tables, and structure from any document

ParameterRequiredTypeDescription
Urlsource urlSourceYestext
Base64Source base64SourceYestext

Tips

  • OCR + table detection + paragraph + heading structure. Good for general structured extraction.

OCR Read

Simple OCR – extract text from images/PDFs

ParameterRequiredTypeDescription
Urlsource urlSourceYestext
Base64Source base64SourceYestext

Tips

  • Plain OCR – text only. Cheapest model. Use for indexing / searchable archives.
  • urlSource: publicly accessible URL OR base64Source: base64-encoded bytes (max 4 MB).

Analyze Receipt

Extract data from a receipt

ParameterRequiredTypeDescription
Urlsource urlSourceYestext
Base64Source base64SourceYestext

Tips

  • Normalised: MerchantName, TransactionDate, Items[], Total. Optimised for printed receipts.

Get Analysis Result

Get result of an async analysis operation

ParameterRequiredTypeDescription
Model Id model_idYestext
Result Id result_idYestext

Tips

  • Multi-page docs may run async – Start operations return an Operation-Location. Use Get Result to poll status.

List Models

List available document models

No input parameters.

Tips

  • Returns prebuilt + custom models available in your resource. Use to discover custom-model IDs.

FAQ

Doc Intelligence vs AWS Textract vs Google Doc AI – which?
Comparable accuracy on common docs. Doc Intelligence's invoice/receipt models are slightly more business-process-friendly out of the box. Textract has stronger forms/tables on hand-written docs. Test all three on your representative docs.
Custom model training – how?
In Azure Document Intelligence Studio. Upload ~5 labelled samples per field, the service trains a model + returns the model_id. Then call Analyze Custom with that ID.
URL vs base64 input?
URL is faster (Azure fetches directly). Base64 is needed for files not publicly hosted – max 4 MB request size. For larger files, push to Blob Storage first + pass the URL.
v3 vs v4 API?
Connector uses v4 (2024-11-30 GA). v3 is supported for backwards compat. New builds should use v4.

Related connectors

See it working on your own data

Everything documented here ships with the platform – try the document tools free, or go live in 7 days.