Newly Added AI Models in Power Platform

In this blog, we will provide an overview of a few newly introduced AI models within the Power Platform. These models include the new Entity extraction prebuilt model, Entity extraction custom model, and Category classification models.  

Here is an overview of these models: 

1. Entity Extraction an AI Model  

This model recognizes specific data (textual) that is essential and interesting for the business. The model uses predefined categories to identify the elements from the text and then classifies them into the appropriate categories. This model is helpful for the transformation of unstructured data into machine-readable data. An example of a use for this model would be to extract and classify the data in a string of text. Parsing the text “Costs have increased by 7% at our Carlsbad office” and language “English” Would provide and output of: “7%” as a Percentage and “Carlsbad” as a city. 

For exploring this model, we can use the “try it out” feature provided in the Power Apps portal.  

Follow these steps to explore the model: 

  1. Go to the make. PowerApps portal and select AI Builder à Build.  
  2. Then go to Entity Extraction and select the “Try it out” option.  
  3. Use predefined text samples to analyze. 

This model does have some limitations: 

  1. The length of the document is a Maximum of 5000 Characters. 
  2. The supported languages are limited to: English, Chinese-Simplified, French, German, Portuguese, Italian and Spanish. 
  3. Supported entity types are Age, Boolean, Color, Continent, Date and time and a few others.  

2. Entity Extraction Custom AI Model  

This model is the Entity Extraction model discussed above. It recognizes the specific data from the text and structures it with the help of predefined categories to classify it in informative data essential/useful for the business.  

So, what is the difference between these two models?  

This model requires you to build, extract and publish your Custom entity before extracting and using it. As you make/construct data and design parameters, you also create the entity extraction model with a unique requirement for your model.   

3. Category Classification Custom AI Model  

One primary problem for many organizations is the volume of data coming from different channels, such as email, social media, and documents. This data might have valuable information that can be very helpful for a business if analyzed quickly. The primary hurdle is to extract/categorize the data carried in that volume of data cost-effectively.  

The Category Classification model is one of the fundamental natural language processing models. We can identify tags in the text for sentiment analysis, spam detection, and customer request routing. 

Empower Your Business: Learn About Power Platform!

Learn More

Example of a Category Classification Custom Model: 

For this final example, let’s create a Category Classification Custom Model in Power Platform by following the steps mentioned below: 

Step # 1: Log in to the portal. 

Step # 2: Click on the AI Builder> Build > Category Classification. The process is highlighted in the image below. 

Click on the AI Builder

Step # 3: Give a name to your AI model and then click ‘Create’ button. 

Give a name to your AI model and then click 'Create' button

Step # 4: Now, a wizard has started. We need to click on the ‘Select text’ button, and it will give us the list of tables in the Dataverse. 

Select text

Step # 5: After selecting the table, we need to choose the option “Account Number” and then click the ‘Select’ column. After that, you will get the preview of data and then press the ‘Next’ button at the bottom of the screen. 

selecting the table and then select Account Number

Step # 6: Now, you have to select the tags. Click on the ‘Select tags’ button, and then it will show the list of columns for the same table chosen in the previous step. In this step, we are selecting “Account Name” as a tag. After that, click on the select column, it will ask for the separator. If there is any in your text, then select “Open.” Otherwise, click “Next.” 

Account Name

Step # 7: Now, it will give you an overall review. Check it and then click “Next.” 

Check it and then click

Step # 8: Select the language for this example. We are using “English” and then click “Next.” It will give you a model summary. After that, click “Train” at the bottom of the wizard. 


Step # 9: Your model is created. We can further train and finalize it. 

Your model is created. We can further train and finalize it


This blog gives us a general idea of the AI model for data extraction and categorizing it. These models help organize the unstructured data in a structured one with the use of tags and categories.