How to Use Azure AI Document Intelligence in Power Automate Flows

In the world of automation and AI, one of the trickiest parts has always been dealing with documents — PDFs, scanned files, invoices, forms, and whatnot. Manually extracting data from these files is not just boring but also error prone. That’s where Azure AI Document Intelligence comes in, and when we combine it with Power Automate, the results are just brilliant.

In this article, I’ll Walk you through what Azure AI Document Intelligence is, how it can be helpful, and how you can integrate it with Power Automate to build smart document workflows — all without writing any complex code.

📄 What is Azure AI Document Intelligence?

Azure AI Document Intelligence (formerly called Form Recognizer) is a cloud-based AI service from Microsoft that uses machine learning models to extract structured data from unstructured or semi-structured documents.

Think of it like this — you upload a receipt, a medical form, or an invoice, and Document Intelligence understands the document layout, detects key-value pairs, tables, and even handwritten content, and then gives you structured JSON output. It’s like giving your documents eyes and a brain!

🤔 Why Document Intelligence is a Game-Changer?

Before tools like Document Intelligence, companies had to rely on manual data entry or build custom OCR logic using outdated tools. This used to be:

• Time-consuming
• Prone to human error
• Not scalable
• Difficult to maintain

Also, working with scanned documents or complex formats like tax forms, resumes, or delivery notes meant writing hundreds of lines of parsing logic, regular expressions, and dealing with document variations.

With Document Intelligence, you just upload the document, and you get neatly extracted data — almost like magic!

💼 Real-Time Use Cases

Here are a few real-world examples where Document Intelligence really shines:

1. Accounts Payable Automation: Automatically extract invoice numbers, due dates, total amounts, and vendor names from PDF invoices.
2. Loan Application Processing: Read bank statements, income proof, and ID cards from customer-submitted documents.
3. Healthcare Automation: Pull diagnosis codes, lab values, and patient info from lab reports or handwritten doctor notes.
4. Shipping and Logistics: Extract delivery note information, pickup receipts, and package details from scanned docs.
5. Resume Parsing: Auto-extract candidate name, skills, and work experience from resume PDFs and feed into HR systems.

🔧 How to Use Azure Document Intelligence in Power Automate?

Let’s now move to the fun part, integrating Document Intelligence with Power Automate.

🔹 Step 1. Set Up the Azure Document Intelligence Resource

Here’s a full guide to get your Azure resource ready:

✅ 1. Log in to Azure

Go to https://portal.azure.com and sign in with your Azure credentials.

✅ 2. Search for “Document Intelligence”

Type “Document Intelligence” or “Form Recognizer” in the top search bar and select the service from the Marketplace.

✅ 3. Click “Create”

Click “+ Create” to begin setting up the resource.

✅ 4. Fill in Basic Details

Field	Value (Example)
Subscription	Your Azure Subscription
Resource Group	doc-ai-rg (new or existing)
Region	Choose closest (e.g., East US, Central India)
Name	doc-intel-demo
Pricing Tier	Free F0 (for testing) or Standard S0 (for production)

Click Review + create, then Create.

Deployment in progress,

✅ 5. Go to the Resource

Once deployment finishes, click “Go to resource”.

Click AzureDocumentIntelligence

✅ 6. Get API Key and Endpoint

• Endpoint: Available under the Overview tab — e.g: https://azuredocumentinteligence.cognitiveservices.azure.com/
• API Key: Click “Keys and Endpoint” from the left menu. Copy Key1 or Key2.

⚠️ Keep these confidential. These will be used inside Power Automate's HTTP request.

✅ Example URL: https://azuredocumentreaderbygowtham.cognitiveservices.azure.com/formrecognizer/documentModels/prebuilt-invoice:analyze?api-version=2023-07-31

🔍 Breakdown of Each Section

Part	Explanation
https://azuredocumentreaderbygowtham.cognitiveservices.azure.com	This is your Azure endpoint for the Document Intelligence resource. It is unique to your Azure instance.
/formrecognizer	This indicates you're using the Form Recognizer service.
/documentModels	This points to the document models API, used for both prebuilt and custom models.
/prebuilt-invoice:analyze	You're calling the prebuilt-invoice model and triggering the analyze operation.
?api-version=2023-07-31	Specifies the API version you’re using. In this case, version 2023-07-31. It's important to use the correct version supported by the endpoint and SDK.

✅ 7. Optional – Try Document Intelligence Studio

You can test documents directly using the Studio at: 👉 https://documentintelligence.ai.azure.com/studio

For testing, click: Prebuild invoice model.

Click continue.

For testing, click Run Analysis.

🔹 Step 2. Create Your Power Automate Flow

1. Go to https://make.powerautomate.com
2. Start with a trigger like:
   • When a file is created in a SharePoint folder
   • When an email with an attachment arrives
3. Add an HTTP Action next.

🔹 Step 3. Add HTTP Action to Call Document Intelligence

Configure the HTTP action like this:

• Method: POST

URL: POST

https://azuredocumentreaderbygowtham.cognitiveservices.azure.com/formrecognizer/documentModels/prebuilt-invoice:analyze?api-version=2023-07-31

• Headers:

Content-Type: application/pdf

Ocp-Apim-Subscription-Key: <your-api-key>

• Body: Pass the file content (binary) from the trigger step.

{
    "modelId": "prebuilt-invoice",
    "urlSource": "https://github.com/Azure-Samples/cognitive-services-REST-api-samples/raw/master/curl/form-recognizer/rest-api/invoice.pdf"
}

You can also use a custom model trained via Document Intelligence Studio.

🔹 Step 4. Get the content from the PDF after we post it in Document Intelligence

Compose: Get operation-location

After the POST request, Azure AI doesn't give the result immediately. It gives you an operation URL to fetch the result after processing.

So add a Compose step with:

outputs('HTTP)?['headers']?['operation-location']

Change HTTP to match your POST action name.

Delay for 15 Seconds

Azure needs time to process.

HTTP GET to Fetch the Results

• Method: GET

URI: Use the Compose output directly

@outputs('Compose')

Headers

{
  "Ocp-Apim-Subscription-Key": "<your-key>"
}

Parse JSON: Get the Final Output

Content

body('HTTP_2')

Schema for prebuilt-invoice (for Parse JSON)

Use this as your schema (basic structure):

Sample parse Json schema attached in this article. Please use it.

@{body('Parse_JSON')?['analyzeResult']?['documents']?[0]?['fields']?['InvoiceId']?['valueString']}

Overall flow,

Output

Set variable is

@{body('Parse_JSON')?['analyzeResult']?['documents']?[0]?['fields']?['InvoiceId']?['valueString']}

🔹 Step 5. Add Post-Actions

Based on the extracted content, you can:

• Create a SharePoint item
• Send an approval in Teams
• Store data in Dataverse
• Trigger custom workflows

✅ Key Benefits of This Integration

• Zero manual work: automate document reading
• Highly accurate: thanks to AI models
• Flexible integration: works with many triggers (email, SharePoint, etc.)
• Low-code: no need to be a full-time developer

🧠 Final Thoughts

Every business works with documents, and most of them still involve manual reading or entry. With Azure Document Intelligence and Power Automate, you can remove this pain entirely.

This combo is perfect for finance, HR, logistics, healthcare, or any domain that deals with forms, receipts, or applications. And the best part? It’s fast, low-code, and super scalable.

📌 Quick Recap

• Use Azure Document Intelligence to extract data from documents like invoices, receipts, and forms.
• Set up the service in Azure and get your API Key + Endpoint.
• Build a Power Automate flow using an HTTP action to call the API.
• Parse the result and trigger actions like approval or record creation.

So go ahead, try it out and start building smarter workflows!

Conclusion

Let me know in the comments if you'd like me to write about training custom models or building multi-document batch flows!

Happy automating! 🛠️