Use the Payloads: Run Payload Flow action to start a CSV import from a Salesforce File. The action accepts and queues the work; the file is processed after the starting transaction commits.

Configure a CSV Import Payload and its Salesforce file connection. Grant the actual Flow execution identity access to the connection principal, the file and Payloads runtime operations.

The file connection is mandatory even for a screen Flow. See Set up the Salesforce file connection.

Selecting a CSV Import Payload reveals CSV File ID and the cleanup option. This configuration example leaves the file input blank and shows the missing-connection warning; it is not a completed Flow import.

ContentVersion IDs begin with 068; ContentDocument IDs begin with 069. A document resolves to its latest accessible version at acceptance, and that exact version remains pinned for the import.

A screen file-upload output is a collection. Select one file ID for the action rather than mapping the whole collection into CSV File ID.

Dynamic Inputs are saved for use in every chunk. Rows per chunk comes from the CSV Import Action settings; the general Batch Size setting does not control CSV chunks.

Add a Decision after the action. If success is true, retain the returned Job ID and continue with any acceptance acknowledgement. If success is false, use error to handle the invalid request.

This normal error response is not the Flow fault connector. Keep a fault path for unexpected Flow failures, but do not rely on it to handle success being false.

Success means accepted and queued, not processed. Check Job status and row outcomes in a later transaction. Do not poll for completion in the starting Flow transaction.

The Flow can create or link the file before invoking the action in the same transaction. Workers start after commit; rolling back that transaction also rolls back the newly created Jobs and dispatch request.

A bulk invocation returns one response per input, in input order. Invalid requests do not prevent valid requests in that invocation from being accepted. Multiple imports use a dispatcher and consume native asynchronous capacity.

Calling the action repeatedly inside a Flow loop creates separate invocations and still consumes the transaction's asynchronous limits. Do not treat the bulk dispatcher as permission for an unbounded loop.

Source cleanup can delete the whole File after completion, including completion with row errors. Leave it disabled if another import or business process needs the File.

Test the actual screen, record-triggered or unattended Flow context. Include invalid file access and row failures, then inspect CSV import results.