Considerations for Complaint Intake with Agentforce

Complaint Categorization (Type and Subtype): From a constituent’s description, the agent determines the type and subtype for the complaint by selecting from the allowed picklist values on the Public Complaint object. To ensure that complaints are categorized correctly:

• Review and customize the picklist values for the Type and Subtype fields on the Public Complaint object to match the kinds of issues your agency handles.
• Deactivate any predefined values that don’t apply. See Add or Edit Picklist Values.
• After you customize the picklist values, update the instructions in the Complaint Intake subagent to guide the agent’s selections. See Edit a Standard Subagent in the Legacy Builder.

Handling Sensitive Complaints and Escalation: By default, the Einstein Trust Layer scans prompt responses for toxic language. However, to effectively manage sensitive complaints, you must enhance the agent's default behavior. The agent includes a standard Inappropriate_Content subagent designed to stop the process if it detects sensitive or inappropriate user input (such as violence, harassment, illegal activities, and so on). This default behavior is not sufficient for a human handoff. To ensure that constituents receive proper assistance, you must override the Inappropriate_Content subagent with new behavior that calls the Escalation subagent to transfer the conversation to a human agent. See Create the Complaint Intake Agent for instructions.

Email Acknowledgment Limits: The Email Complaint Acknowledgment flow uses the Send Email flow action. Be aware that each licensed org can send single emails to a maximum of 5,000 external email addresses per day (based on GMT). See Flow Core Action: Send Email.