How to read file content using Salesforce Agentforce

For example, suppose you want to extract:
o    First Name
o    Last Name
o    Date of Birth
o    Address

You can use a prompt similar to the following:
You are a document classification and extraction agent.
Analyze the attached file and identify:
o    FirstName
o    Lastname
o    Birthdate
o    Address
Return only the following JSON:
{
 "firstname": "",
 "lastname": "",
 "birthDate": "",
 "address": ""
}

Rules:
o    Extract only values found in the document.
o    Do not guess missing values.
o    Return empty strings if data is unavailable.
o    Return only JSON.
o    Keep the address in a single line.

A well-written prompt significantly improves extraction accuracy.

Step 4 : Upload a Document
7. Once the Prompt Template is configured:
    Open Preview Settings.
8. Upload a sample document. 
9. Click Add.
    Agentforce will use the uploaded document as the input for AI processing.
    Sample document: 

Step 5 : Generate Structured JSON
10. Click Save & Preview.
11. Agentforce analyzes the document and returns structured JSON similar to:
Example:
         {
           "firstname": "Michael",
           "lastname": "Johnson",
           "birthDate": "April 12, 1990",
           "address": "123 Maple Street San Diego, CA 92101 United States"
         }
        Returning JSON instead of plain text makes it easy for Apex and Salesforce
        integrations to  consume the extracted data.
Step 6 : Process the JSON in Apex
12. After Agentforce extracts the document information, the JSON response is          sent to an Apex class.
The Apex class performs the following steps:
o    Receives the JSON response.
o    Deserializes the JSON using JSON.deserialize() or
      JSON.deserializeUntyped().
o    Extracts the required field values.
o    Performs business validations.
o    Creates or updates Salesforce records.
o    Returns the appropriate response back to Agentforce.
Depending on your business requirements, the extracted information can be used to:
o    Validate records.
o    Create new records.
o    Update existing records.
o    Trigger additional automation.

Step 7 : Read File Content in Apex
13. Salesforce stores uploaded files using the ContentVersion object. 
To access a file:
a.   Receive the ContentVersion Id.
b.   Query the required ContentVersion record.
c.   Retrieve fields such as:
o    Id
o    Title
o    FileExtension
o    VersionData
o    ContentDocumentId
o    FileType
d.   Read the VersionData Blob.
14. The Blob can then be converted depending on the file type:
o    Text files → String
o    CSV files → Parse rows
o    JSON files → Deserialize objects
15. Once the content is available, Apex can execute any required business logic,
      such as validating data or comparing it with Salesforce records.

Benefits of This Approach
Using Agentforce together with Apex provides several advantages:
o    Reduces manual data entry.
o    Extracts structured information from documents.
o    Improves accuracy through AI-powered extraction.
o    Keeps AI processing separate from business logic.
o    Makes the solution reusable and easy to maintain.
o    Integrates seamlessly with Salesforce automation.

Conclusion
Salesforce Agentforce makes document processing significantly easier by combining generative AI with Salesforce automation. Agentforce understands and extracts information from uploaded documents, while Apex focuses on validation, business logic, and Salesforce integration.
This separation of responsibilities creates a scalable, maintainable, and reusable solution for document-driven business processes. Whether you're processing identity documents, invoices, contracts, or application forms, this approach helps automate workflows while reducing manual effort and improving data quality.

Detailed Instruction: Read file content using Salesforce Agentforce