#Agentforce

Capacity model + a custom skill for 'AE-routable-deal' is what's worked for us. We don't keep the agent in the conversation mid-stream — it posts a structured handoff summary as the first message in the new conversation, and the AE picks it up cold with the full context attached. Cleaner than 'silent watching'.

Two-stage retrieval has been the cleanest pattern: pull the case history through a redaction layer first (the Einstein Trust Layer handles a lot of this by default but you can extend it), then ground the agent prompt on the redacted versions. Your compliance team should be able to approve the redaction policy once, not every prompt.