Intelligent Form Reader setup, document templates and field mapping
Key Takeaway
Create a document type and a template for each kind of document, then map the template's document fields to fields on one target object. In Agentforce Financial Services, only the Default record type is supported for storing extracted data. Keep Amazon Textract Analyze ID turned off because it isn't compatible.
Intelligent Form Reader uses optical character recognition (OCR) through Amazon Textract to extract data from financial documents, such as W2 forms and tax returns. You specify the source form, then map the fields in that form to the equivalent fields in Salesforce. The extracted data can create or update record fields. It can also verify data that's already in the org. For example, a banker can check a client's birthdate in Salesforce against the birthdate in a scanned passport.
How document extraction works
Users add submitted documents from the Received Documents view. The process begins when a user uploads a PDF, JPG, or PNG file. Agentforce Financial Services then creates a record for the uploaded document.
A Received Document record is created every time a user uploads an image file and manipulates the uploaded file. By default, the Received Document has the same name as the uploaded file. The user can select the pages to extract data from and discard the rest. The user can also rotate pages if needed.
File Rendition services change the uploaded document into an image. The Amazon Textract API then processes that image to extract the data, as a background OCR process. The extracted data is mapped to Salesforce objects as the admin defined.
The rendition and the Amazon Textract extraction are asynchronous, so extraction can take some time. If extraction takes longer than 50 seconds, the user can retry by uploading the file again. If OCR extraction fails, resend the document.