How does ChatNexus detect when to escalate to a human?
Direct answer:
ChatNexus monitors conversation confidence scores and user intent. When the agent confidence drops below a threshold (e.g., the agent says 'I'm not sure') or the user explicitly requests a human ('Can I talk to someone?'), the system triggers escalation.
The escalation is automatic for explicit requests and threshold-based for ambiguous cases. You configure the confidence threshold in Agent Settings → Escalation Rules.
What happens when a conversation is escalated?
Direct answer:
The chat moves to the <strong>Operator Inbox</strong>, where your team receives a live notification with full conversation history. An operator can reply immediately, and the conversation resumes in the chat widget.
The conversation context—user's original query, all previous AI responses, and any file uploads—are preserved. The operator sees everything and can respond in the same thread. The user sees responses in real-time via WebSockets (Laravel Reverb).
Can I set up escalation rules for specific topics?
Direct answer:
Yes. You can create rules like 'if topic is billing, escalate immediately' or 'if the user mentions their account number twice, escalate.' These are configured in the agent dashboard.
Rules can be based on keywords, conversation length, AI confidence, or user intent tags. You can also set up automatic routing—for example, 'route billing queries to the billing team' using tags.
How are operators notified about incoming escalations?
Direct answer:
Operators see live notifications in the ChatNexus operator dashboard via WebSockets. Notifications include the user's name, query preview, and a direct link to the conversation.
Operators can also opt into email or Slack notifications for high-priority escalations. Set up Slack integration in Settings → Integrations → Slack.
Can I re-route a conversation back to the AI after a human responds?
Direct answer:
No. Once a conversation is escalated to a human, the AI is no longer involved. This ensures continuity and prevents the AI from interrupting. The human can choose when to close the conversation or mark it as handled.
If you want the AI to handle follow-ups after the human provides context, the operator can manually trigger AI resumption via a button in the chat. However, the default is human-only to avoid confusion.
Deploy an AI agent on your site in minutes
Connect your knowledge base, embed on any domain, and start free — no credit card required.
Create Your AI Agent Free