How To Extract Tables From A PDF In Power Automate

Imported reference from Matthew Devaney's blog for learning purposes. Original: https://www.matthewdevaney.com/how-to-extract-tables-from-a-pdf-in-power-automate/. Author: Matthew Devaney.

How To Extract Tables From A PDF In Power Automate Posted by - Matthew Devaney on - May 4, 2025 7 Comments

Power Automate can extract tables from a PDF with the help of Azure AI Document Intelligence . Using the pre-built Layout model we can analyze any PDF document and get its information in a JSON. The output includes an array containing all of the tables in the PDF. Table of Contents

“>Introduction: The Extract Tables From A PDF Automation

Create A New Document Intelligence Resource

Locate The Document Intelligence Resource Keys and Endpoint

Try The Layout AI Model In Document Intelligence Studio

Process The PDF Document With The Layout AI Model

Get The Document Intelligence Layout AI Model Results

Extract The Table Cells From The Layout Model Results

Run The Power Automate Flow To Extract Tables From A PDF View The Completed Power Automate Flow

Introduction: The Extract Tables From A PDF Automation

An automation connected to Document Intelligence accepts a pdf file as input and extracts all tables found within in the document. No training of the AI model is needed in advance for new document types.

Tables can be extracted over multiple pages. There can be multiple tables on a single page.

The output is an array of values with a row number, a column number and the content of the table cell.

Create A New Document Intelligence Resource

The first thing we must do is create a new document intelligence resource. Open the Azure Portal and search for Document Intelligences .

Press the Create button on the document intelligence menu.

Name the document intelligence resource LayoutAnalyzerAITutorial and select a pricing tier of Standard S0 or above. Then press the review + create button to complete the setup.

Locate The Document Intelligence Resource Keys and Endpoint

We will need the document intelligence resource’s keys and endpoint when building our automation. Navigate to the Keys and Endpoint menu item within the Resource Management grouping.

When we need the keys and endpoint later on come back to this menu to find them.

Try The Layout AI Model In Document Intelligence Studio

Before we build an automation to extract all tables from a PDF we can try out the AI model in Document Intelligence Studio. It allows us to manually upload a PDF document and see what the results will look like. Go to the Overview menu and press the Go To Document Intelligence Studio button. Select the Layout model.

Update the service resource with your keys and endpoint for the document intelligence resource we just created.

Drag and drop a file into the upload area and select the Run Analysis button.

After the analysis is run we can see the tables that were extracted from the PDF by the AI model. The image below shows only 1 table for the first page. But if we navigate to page 2 of the document we will see 2 more tables.

Process The PDF Document With The Layout AI Model

We are ready to build the Power Automate flow to call Document Intelligence Layout model. Open Power Automate and create a new instant flow named Extract Tables From Document .

Add a file input to the flow trigger named File Content .

Then insert and HTTP action into the flow to send the file to the Layout AI model. Use the POST method. POST

Copy the endpoint from the Keys and Endpoint menu in Azure document intelligence. Then specify the prebuilt layout model and choose the api version .

https: //layoutanalyzeraitutorial.cognitiveservices.azure.com/documentIntelligence/documentModels/prebuilt-layout:analyze?api-version=2024-11-30 Code language: JavaScript ( javascript )

Authorize the flow to use the Document Intelligence service by supplying the API key in the Headers. Header Key Header Value Ocp-Apim-Subscription-Key <Document Intelligence Key>

And pass the PDF file to Document Intelligence in the body of the HTTP action. {

"base64Source" : @{triggerBody()?[ 'file' ]?[ 'contentBytes' ]} } Code language: JavaScript ( javascript )

When testing the HTTP Post action it will return a status code of 202. This is because it accepts the request to analyze the document asynchronously. It does not immediately return the AI processing results.

Get The Document Intelligence Layout AI Model Results

Now we must perform another set of actions to get the Layout AI model results. Initialize a variable named varApimRequestId and choose a string data type. The variable will store the unique identifier of the batch we want to get the results from.

Use this code in the Value property of the action.

outputs( 'HTTP:_POST_Layout_Model' )[ 'headers' ]?[ 'apim-request-id' ] Code language: JavaScript ( javascript )

Add a 10-second delay to the flow. If the AI processing is taking longer than 10 seconds we might need to increase the count.

Insert another HTTP action into the flow to get the results from the Document Intelligence Layout AI model.

Make the HTTP request using the GET method. GET

Start with the same endpoint as the HTTP Post action but modify it to include the varApimRequestId variable which specifies the batch.

https: //layoutanalyzeraitutorial.cognitiveservices.azure.com/documentIntelligence/documentModels/prebuilt-layout/analyzeResults/@{variables('varApimRequestId')}?api-version=2024-11-30 Code language: JavaScript ( javascript )

Choose a the application/json Content-Type and include the API key once again. Header Key Header Value Content-Type application/json Ocp-Apim-Subscription-Key <Document Intelligence Key>

When we give the flow a test run it waits 10-seconds and then retrieves the AI processing results.

Extract The Table Cells From The Layout Model Results

The Document Intelligence layout model results contain more information than we need. We only want to get the tables from the document. Therefore, we must extract the tables from the AI model results.

Add a Compose action to the flow and use this flow expression to get all of the tables in an array.

body( 'HTTP:_GET_Layout_Model_Result' )?[ 'analyzeResult' ]?[ 'tables' ] Code language: JavaScript ( javascript )

Each table in the array contains records with information about its cells: a row number, a column number and the cell contents. Add a Select action to the flow to get this information.

Use this code in the Select action to get the row number, the column number and the cell contents. Key Value columnIndex item()?[‘columnIndex’] content item()?[‘content’] rowIndex item()?[‘rowIndex’]

Run The Power Automate Flow To Extract Tables From A PDF

We are finished building the Power Automate flow. Run the flow to make sure it works.

The flow outputs an array for each table with the row number, column number and contents of each cell. View The Completed Power Automate Flow

Here is a screenshot of the full Power Automate flow. Did You Enjoy This Article? 😺

Subscribe to get new Copilot Studio articles sent to your inbox each week for FREE Enter your email address Sign Me Up Questions?

If you have any questions or feedback about How To Extract Tables From A PDF In Power Automate please leave a message in the comments section below. You can post using your email address and are not required to create an account to join the discussion. Matthew Devaney Power Automate

Save Docusign Document To SharePoint With Power Automate