Azure Document Intelligence: Automating document processing with AI

Azure Document Intelligence helps teams automate the extraction of text, tables, key-value pairs, and structured fields from documents such as invoices, receipts, contracts, identity documents, and business forms.

Instead of manually reading files and entering data into systems, enterprises can use AI document processing to turn documents into structured outputs that applications and workflows can use.

It is especially useful for automated document processing, document data extraction, invoice data extraction, and building end-to-end workflows with Azure Document Intelligence Power Automate integration.

What is Azure Document Intelligence?

Azure AI Document Intelligence is a Microsoft Azure AI service used for intelligent document processing. It applies OCR, machine learning, and document understanding models to analyze documents and extract useful information.

The service was previously known as Azure Form Recognizer, so you may still see that name in older documentation, connectors, or tutorials. Power Automate in the Microsoft connector also references Azure AI Document Intelligence, formerly Form Recognizer.

With Azure Document Intelligence, users can submit documents such as PDFs, images, scanned files, or forms. The service analyzes the content and returns structured data, often in JSON format, that can be used in business applications, databases, approval workflows, or reporting systems.

Why does Azure Document Intelligence matter?

Many business processes still depend on manual data entry. Teams often need to open documents, read values, copy information, validate fields, and move data into another system.

This creates common problems:

Manual document processing issue	Business impact
Slow data entry	Delays in finance, HR, legal, and operations
Human error	Incorrect invoice totals, names, dates, or IDs
Poor scalability	More documents require more manual effort
Unstructured data	Hard to search, report, or automate
Repetitive tasks	Employees spend time on low-value work

Azure Document Intelligence solves this by supporting OCR document processing, data extraction, and automation. It helps convert unstructured or semi-structured documents into usable business data.

Read more: Unlocking the hidden value of Microsoft Azure AI: An analysis of its total economic impact

How does Azure Document Intelligence work?

At a high level, the process works like this:

1. A document is uploaded to Azure AI Document Intelligence.
2. The selected model analyzes the document.
3. The service detects text, layout, tables, fields, and key-value pairs.
4. The extracted result is returned as structured data.
5. The output can be sent to apps, databases, SharePoint, Dataverse, Teams, or Power Automate workflows.

This makes the service useful for document data extraction, AI invoice processing, customer onboarding, contract review, expense processing, and compliance workflows.

Take control of your business operations

Discover how Confiz services can simplify your complex workflows and improve decision-making.

Get a Free Quote

Main model types in Azure Document Intelligence

Azure Document Intelligence offers various model options based on the document type and the level of customization required. Microsoft supports prebuilt domain models and custom models for business-specific use cases.

Model type	Best for	Training required?
Document analysis models	General text, layout, tables, and key-value extraction	No
Azure Document Intelligence prebuilt models	Common documents such as invoices, receipts, IDs, contracts, and business cards	No
Azure Document Intelligence custom models	Unique internal forms, company-specific templates, or special document formats	Yes

1. Document Analysis Models

Document analysis models are general-purpose models that understand the structure and content of documents.

They can detect:

• Text
• Tables
• Selection marks
• Key-value pairs
• Paragraphs
• Layout elements
• Document structure

These models are useful when the document type is unknown or when you only need general OCR document processing and layout extraction.

For example, if you receive mixed PDFs with tables, paragraphs, and labels, a document analysis model can help extract the basic structure before additional processing happens.

2. Azure Document Intelligence Prebuilt Models

Azure Document Intelligence prebuilt models are ready-to-use models trained for common business document types. These models do not require manual training or custom labeling.

You send the document to the relevant prebuilt model, and the service returns structured fields.

Below are the 13 major prebuilt models.

1. Invoice Model

The Invoice Model extracts structured data from vendor and supplier invoices. It is one of the most common use cases for invoice data extraction, AI invoice processing, and Azure invoice processing.

It can extract fields such as:

• Invoice number
• Vendor name
• Vendor address
• Customer details
• Invoice date
• Due date
• Subtotal
• Tax amount
• Total amount
• Line items
• Product descriptions
• Quantity
• Unit price

Common use case

Businesses use the invoice model to automate accounts payable processes, reduce manual data entry, and improve financial accuracy.

2. Receipt Model

The Receipt Model extracts information from retail purchase receipts.

It can extract:

• Merchant name
• Transaction date
• Transaction time
• Purchased items
• Tax amount
• Total amount paid

Common use case

Organizations use this model for expense management, employee reimbursements, and receipt-based approval workflows.

3. Identity Document Model

The Identity Document Model extracts personal details from identity documents such as passports, driver’s licenses, and national ID cards.

It can extract:

• Full name
• Date of birth
• Gender
• Document number
• Expiry date
• Nationality

Common use case

This model is useful for onboarding, KYC processes, banking verification, identity checks, and customer registration workflows.

4. US Health Insurance Card Model

The US Health Insurance Card Model extracts information from healthcare insurance cards.

It can extract:

• Member name
• Insurance provider
• Prescription information
• Group number
• Policy number

Common use case

Healthcare providers can use this model to automatically capture patient insurance details during registration or medical processing.

5. US Personal Tax Model

The US Personal Tax Model processes tax documents such as W-2 and 1099 forms.

It can extract:

• Employer details
• Employee information
• Income amounts
• Tax deductions
• Tax identification numbers

Common use case

Accounting firms, tax services, and finance teams can use this model to automate tax document processing for powerful automation in finance.

6. US Mortgage Documents Model

The US Mortgage Documents Model processes mortgage and property financing documents.

It can extract:

• Borrower information
• Property details
• Loan amount
• Interest rate
• Mortgage terms

Common use case

Banks, lenders, and financial institutions use this model to simplify loan application processing and mortgage document review.

7. US Pay Stubs Model

The US Pay Stubs Model extracts payroll information from employee pay slips.

It can extract:

• Employee name
• Employer name
• Payment period
• Gross earnings
• Deductions
• Net pay

Common use case

HR, payroll, and financial verification teams use this model for income verification and salary documentation.

8. US Bank Statement Model

The US Bank Statement Model extracts information from bank-issued financial statements.

It can extract:

• Account holder name
• Bank name
• Account number
• Transaction dates
• Deposits
• Withdrawals
• Account balance

Common use case

This model is helpful for financial analysis, auditing, loan verification, and customer financial review.

9. US Checks Model

The US Checks Model extracts information from bank checks.

It can extract:

• Payee name
• Date
• Amount
• Bank routing number
• Check number

Common use case

Banks and financial institutions use this model to automate check processing and verification.

10. Credit Card Model

The Credit Card Model extracts key information from credit card images or documents.

It can extract:

• Cardholder name
• Card number
• Expiration date
• Issuing bank

Common use case

This model can support payment verification and financial record processing where card details need to be captured securely.

11. US Marriage Certificate Model

The US Marriage Certificate Model extracts information from official marriage documents.

It can extract:

• Names of spouses
• Date of marriage
• Location of marriage
• Names of witnesses or officials

Common use case

Government agencies and record management teams can use this model for document digitization and official record processing.

12. Contract Model

The Contract Model extracts important information from legal agreements and contracts.

It can extract:

• Contract title
• Parties involved
• Agreement dates
• Terms and conditions
• Signatories

Common use case

Legal, procurement, and operations teams use this model for contract management, legal review, and compliance workflows.

13. Business Card Model

The Business Card Model extracts contact information from business cards.

It can extract:

• Person’s name
• Job title
• Company name
• Phone number
• Email address
• Website

Common use case

Sales and marketing teams can use this model to automatically save contact details into CRM systems.

Prebuilt model comparison table

Prebuilt model	Best for	Example fields
Invoice Model	Azure invoice processing	Invoice number, vendor, dates, totals, line items
Receipt Model	Expense claims	Merchant, items, tax, total
Identity Document Model	ID verification	Name, DOB, document number, expiry
US Health Insurance Card Model	Healthcare intake	Member name, provider, group number
US Personal Tax Model	Tax processing	Employer, employee, income, tax IDs
US Mortgage Documents Model	Loan processing	Borrower, property, loan amount
US Pay Stubs Model	Payroll verification	Earnings, deductions, net pay
US Bank Statement Model	Financial review	Transactions, balance, account details
US Checks Model	Check processing	Payee, amount, routing number
Credit Card Model	Payment verification	Cardholder, card number, expiry
US Marriage Certificate Model	Official records	Spouse names, marriage date, location
Contract Model	Contract review	Parties, dates, terms, signatories
Business Card Model	CRM contact capture	Name, company, phone, email

3. Azure Document Intelligence Custom Models

Azure Document Intelligence custom models are used when prebuilt models do not match your document format or business requirements.

Custom models are useful when:

• Your organization uses internal forms
• Your documents have unique layouts
• You need to extract custom fields
• Your documents do not match any prebuilt model
• You process multiple document types
• You need a business-specific extraction workflow

There are two main custom model types: classification models and extraction models.

Accelerate growth at an unprecedented pace

Discover how Confiz can help you take control of your daily operations, increasing growth and revenue.

Book a Free Consultation

Custom Classification Model

A custom classification model identifies and categorizes documents into different types.

For example, a company may receive:

• Invoices
• Purchase orders
• Receipts
• Contracts
• Identity documents

The classification model analyzes the document and determines its type. This is useful when building an automated document-processing workflow in which documents must be routed before extraction.

Example

If a finance team receives mixed documents in one SharePoint folder, the classification model can identify whether a file is an invoice, receipt, or purchase order. Then the workflow can send the file to the correct extraction model.

Custom Extraction Model

A custom extraction model extracts specific fields from documents with known layouts.

It can extract fields such as:

• Names
• Dates
• Identification numbers
• Amounts
• Reference numbers
• Approval status
• Policy numbers
• Internal business fields

This is useful when standard prebuilt models are not enough and the business needs accurate document data extraction from custom templates.

Example

A company may have a custom onboarding form. A custom extraction model can be trained to extract employee name, joining date, department, manager name, employee ID, and approval status.

Prebuilt models vs custom models

Feature	Prebuilt models	Custom models
Training required	No	Yes
Setup time	Faster	Longer
Best for	Common document types	Unique business documents
Flexibility	Limited to supported fields	High
Maintenance	Lower	Medium
Accuracy control	Based on a Microsoft-trained model	Depends on training data and labels
Example	Invoice, receipt, ID, contract	Internal form, custom template, company document
Recommended when	Your document type is supported	Your document has unique fields or layout

Which option is best?

The best option depends on the document type and the business goal.

Scenario	Recommended model
You need to extract text, tables, and layout	Document analysis model
You process vendor invoices	Invoice prebuilt model
You process employee receipts	Receipt prebuilt model
You verify identity documents	Identity Document prebuilt model
You process legal agreements	Contract prebuilt model
You process contact cards	Business Card prebuilt model
You have unique internal forms	Custom extraction model
You receive many document types	Custom classification model
You want a no-code workflow	Azure Document Intelligence Power Automate

For most teams, the best starting point is a prebuilt model. If the prebuilt model does not extract the required fields, then a custom model is the better choice.

Training and testing custom models

Training custom models requires sample documents and field labeling.

Training a classification model

To train a classification model:

1. Upload labeled sample documents for each document category.
2. Use Azure Document Intelligence Studio to prepare the training data.
3. Train the classification model.
4. Test it with new documents.
5. Check whether the model correctly identifies each document type.

Training an extraction model

To train an extraction model:

1. Upload sample documents.
2. Label the fields that need to be extracted.
3. Train the model.
4. Test it using new documents.
5. Review field accuracy.
6. Improve labels or add more samples if needed.

Testing is important because real-world documents may vary in quality, format, layout, and scan clarity.

Connecting Azure Document Intelligence with Power Automate

Azure Document Intelligence Power Automate integration helps organizations create no-code or low-code document automation workflows.

This is useful when documents arrive through:

• SharePoint
• OneDrive
• Outlook email attachments
• Microsoft Teams
• Manual uploads
• Business applications

A Power Automate document-processing flow can send documents to Azure Document Intelligence, receive the extracted fields, and use those fields in the next workflow step.

Step-by-step: Azure Document Intelligence with Power Automate

Step 1: Create a flow

Open Power Automate and create either:

• Automated cloud flow
• Instant cloud flow

Common triggers include:

• When a file is created in SharePoint
• When a file is uploaded to OneDrive
• When an email attachment is received
• When a manual button is selected

Step 2: Add the Azure Document Intelligence connector

Add a new step and search for the Azure AI Document Intelligence connector.

In some environments, you may still see references to Azure Form Recognizer, because that was the previous service name.

Select the document analysis action that matches your use case.

Step 3: Configure the connection

Provide the required connection details:

• Azure endpoint
• API key or authentication method
• Model ID
• Document input

The model ID can refer to either:

• A prebuilt model
• A custom model

Step 4: Provide the document input

Pass the document file content from the trigger into the Azure Document Intelligence action.

Examples include:

• SharePoint file content
• OneDrive file content
• Email attachment content
• Uploaded PDF file
• Uploaded image file

Step 5: Retrieve extracted data

After the document is processed, Power Automate receives structured fields such as:

• Names
• Dates
• Addresses
• Invoice numbers
• Amounts
• Line items
• Table values
• Document IDs

Step 6: Use the extracted data in a workflow

The extracted data can be used to:

• Update SharePoint lists
• Store records in Dataverse
• Save values in Excel
• Send approval requests
• Trigger Outlook emails
• Send Microsoft Teams notifications
• Route documents to departments
• Create audit logs
• Flag documents for manual review

This creates a complete Power Automate document processing solution.

Common business use cases

Use case	How Azure Document Intelligence helps
Invoice processing	Extracts invoice number, vendor, dates, totals, and line items
Expense management	Extracts receipt details for claims and reimbursements
Customer onboarding	Extracts identity document data
Contract management	Extracts parties, dates, terms, and signatories
Healthcare intake	Extracts insurance card information
Loan processing	Extracts mortgage and bank statement details
Payroll verification	Extracts pay stub information
Tax processing	Extracts W-2 and 1099 details
CRM updates	Extracts business card contact information

Benefits of Azure Document Intelligence

Using Azure AI Document Intelligence for intelligent document processing provides several benefits:

• Reduces manual data entry
• Speeds up document handling
• Improves data accuracy
• Supports structured data output
• Works with prebuilt and custom models
• Helps automate business workflows
• Supports scalable automated document processing
• Improves AI document processing use cases
• Enables better OCR document processing
• Works well with Power Automate and Microsoft business tools

Best practices for better results

To get better output from Azure Document Intelligence, follow these best practices:

1. Start with a prebuilt model if your document type is supported.
2. Use a custom model only when prebuilt fields are not enough.
3. Use high-quality PDFs or clear scanned images.
4. Avoid blurry, tilted, or low-resolution documents.
5. Train custom models with real document samples.
6. Label fields consistently.
7. Test models with different document variations.
8. Add validation checks before saving extracted data.
9. Use confidence scores to decide when manual review is needed.
10. Add exception handling in Power Automate.
11. Keep sensitive data handling and access control in mind.
12. Monitor model performance over time.

FAQs

What is Azure Document Intelligence?

Azure Document Intelligence is a Microsoft Azure AI service that extracts text, tables, key-value pairs, and structured fields from documents. It is used for automated document processing, document data extraction, and workflow automation.

Is Azure Document Intelligence the same as Azure Form Recognizer?

Yes. Azure Form Recognizer was the previous name of the service. The current name is Azure AI Document Intelligence.

What is Azure Document Intelligence used for?

It is used for AI document processing, invoice processing, receipt processing, identity verification, contract analysis, tax document processing, bank statement processing, and custom form extraction.

Can Azure Document Intelligence process invoices?

Yes. The invoice prebuilt model supports invoice data extraction and can extract invoice numbers, vendor details, dates, totals, taxes, and line items. This makes it useful for AI invoice processing and Azure invoice processing.

What are Azure Document Intelligence and prebuilt models?

Azure Document Intelligence prebuilt models are ready-to-use models for common business documents such as invoices, receipts, identity documents, health insurance cards, tax forms, mortgage documents, pay stubs, bank statements, checks, credit cards, marriage certificates, contracts, and business cards.

How many prebuilt models are covered in this blog?

This blog covers 13 prebuilt models:

1. Invoice Model
2. Receipt Model
3. Identity Document Model
4. US Health Insurance Card Model
5. US Personal Tax Model
6. US Mortgage Documents Model
7. US Pay Stubs Model
8. US Bank Statement Model
9. US Checks Model
10. Credit Card Model
11. US Marriage Certificate Model
12. Contract Model
13. Business Card Model

What are Azure Document Intelligence and custom models?

Azure Document Intelligence custom models are models trained on your own documents. They are useful when your documents have unique layouts, custom fields, or business-specific formats.

What is the difference between a custom classification model and a custom extraction model?

A custom classification model identifies the document type. A custom extraction model extracts specific fields from the document.

For example, classification can identify a document as an invoice, receipt, or contract. Extraction can then pull values such as invoice number, total amount, contract date, or customer name.

Can Azure Document Intelligence work with Power Automate?

Yes. Azure Document Intelligence Power Automate integration enables users to create automated workflows that analyze documents, extract data, and send results to systems such as SharePoint, Dataverse, Excel, Outlook, or Teams.

What is Power Automate document processing?

Power Automate document processing means using Power Automate to trigger a workflow when a document is received, send it to Azure Document Intelligence, retrieve extracted fields, and use those fields in automated business processes.

Which Azure Document Intelligence model should I use first?

Use a prebuilt model first if your document type is supported. Use a custom model when your document layout is unique or when you need fields that prebuilt models do not extract.

Conclusion

Azure Document Intelligence helps organizations replace manual document handling with scalable intelligent document processing. It supports OCR, machine learning, prebuilt models, custom models, and workflow automation.

For common documents, Azure Document Intelligence prebuilt models provide a fast way to extract structured data. For unique business documents, Azure Document Intelligence custom models offer more control and flexibility.

When combined with Azure Document Intelligence and Power Automate, businesses can build end-to-end automated document-processing workflows that reduce manual effort, improve accuracy, and speed up document-heavy operations.