Loading
Connect in real time with the Trailblazer Community in Slack Open BetaRead More
Automate Your Business Processes with Salesforce Flow
Extract Data from Document Action

Extract Data from Document Action

Extract structured data from documents using AI-powered analysis, with optional human review before the flow continues.

In Flow Builder, add an Action element to your flow. In the field to search actions, enter Extract Data from Document, and select Extract Data from Document.

Set Input Values

Use values from earlier in the flow to set the inputs for the Extract Data from Document action.

Input Parameter Description
Content Document ID The ID of the content document record to analyze.
Document Processing Configuration ID The API name of the document processing configuration used to get data from the document.

Configure Human Review Settings

Use the Human Review Settings section to have a person validate or correct extracted data before the flow continues. When review is required, the flow pauses until the reviewer submits or rejects the review.

Human Review Settings requires Data 360 to be provisioned and enabled in your org.

Input Parameter Description
Let a human review extracted data

Determines when a review task is created. Valid values are:

  • Never — Skip review and return extracted data directly. This value is the default.
  • Only when needed — Create a review task only for extractions where one or more fields fall below the confidence threshold.
  • Always — Create a review task for every extraction, regardless of confidence score.
Assigned To The user who reviews the task.

Store Output Values

To use the extracted data later in the flow, store it in a variable. The values are assigned when the document is processed.

Output Parameter Description
Content Document ID The ID of the Content Document that was processed. This value returns the input Content Document ID.
Document Processing Configuration ID The ID of the IDP configuration used to process the document.
Extracted Data An Apex class containing the extracted data from the document. The class name is determined by the IDP configuration.
Extracted Data JSON A JSON string representation of the extracted data from the document. This output includes an injected needsManualReview flag that indicates whether the extraction confidence score fell below the confidence threshold configured for that field in the document processing configuration.

How Human Review Works

When a review task is created, the flow pauses. Reviewers complete the review in the Human Reviews tab of the Automation app. After the reviewer submits or rejects the task, the flow resumes with the validated or corrected data, or follows the fault path if rejected.

 
Loading
Salesforce Help | Article