Performance may drop due to context handoff issues, unclear transfer triggers, or token limits so make sure assistants are tightly related, transfer triggers are clean and automatic, context isn't lost across calls, and avoid extra prompts before transfers. Also, match voice settings, and reduce token load..