Use a form-processing model in Power Automate

Important

To use AI Builder models in Power Automate, you have to create the flow inside a solution. The steps below won't work if you don't follow these instructions first: Create a flow in a solution.

  1. Sign in to Power Automate, select the My flows tab, and then select New > +Instant-from blank.

  2. Name your flow, select Manually trigger a flow under Choose how to trigger this flow, and then select Create.

  3. Expand Manually trigger a flow, select +Add an input, select File as the input type, and set as input title File Content.

  4. Select + New step, search for AI Builder in the Search for filters and actions box, and then select Process and save information from forms in the list of actions.

  5. Select the form processing model you want to use, select the Document type, and in the Document field add File Content from the trigger:

    Select file content

  6. In the successive actions, you can use any fields and tables extracted by the AI Builder model. For example, let's say that our model is trained to extract the Invoice Id and the Total Amount value, and we want to post those to a Microsoft Teams channel. Just add the Post a message to Teams action, and then select your fields from the list of tokens.

    Note

    • To retreive the value for a field, select <field_name> value . For example, for the INVOICE field, select INVOICE value.
    • To retrieve the confidence score for a field, select <field_name> confidence score . For example, for the INVOICE field, select INVOICE confidence score.

    Form processing flow overview

Parameters

Input

Name Required Type Description Values
AI Model Yes model Form processing model to use for analysis Trained and published form processing models
Document type Yes list The file type of the form to analyze PDF Document (.pdf), JPEG Image (.jpeg), PNG Image (.png)
Form Yes file Form to process

Output

Name Type Description Values
{field} value string The value extracted by the AI model
{field} confidence score float How confident the model is in its prediction Value in the range of 0 to 1. Values close to 1 indicate greater confidence that the extracted value is accurate
{table}{column} value string The value extracted by the AI model for a cell in a table
{table}{column} confidence score float How confident the model is in its prediction Value in the range of 0 to 1. Values close to 1 indicate greater confidence that the extracted cell value is accurate

Note: More output parameters may be proposed such as field coordinates, polygons, bounding boxes and page numbers. These are not listed on purpose as mainly intended to advanced usage.

See also

Overview of the form-processing model