Conversation intelligence and customer satisfaction

klaxify adds a reviewable intelligence layer to every customer conversation. It helps your team understand the request, organize the work, and improve service without hiding what the AI did.

Conversation intelligence

The customer panel can show:

  • a concise conversation summary;
  • the detected customer language and confidence;
  • sentiment and support intent;
  • suggested customer tags;
  • structured information submitted through the website contact form.

Suggested tags are never silently added. Select a suggested tag to apply it to the customer. The original message remains the source of truth.

Duplicate and identity review

When klaxify finds a similar recent conversation for the same customer, it displays a duplicate review card. A workspace manager can link the histories after reviewing the evidence. No messages are deleted or moved, and the link can be undone.

Identity review is separate from duplicate review. It helps a manager reassociate a ticket with a reviewed customer when email or phone details indicate a likely match. Customer records and memory are not destructively combined, and the change can be reversed.

Customer satisfaction surveys

Workspace managers can configure CSAT under Settings → Customer satisfaction surveys. A survey can be prepared whenever a conversation is closed. When Gmail is connected, klaxify sends the survey in the existing email thread. Other channels keep the request pending until a supported delivery path is available.

Customers rate the interaction from 1 to 5 and may add a comment. A score at or below the workspace threshold reopens the conversation and marks it high priority for service recovery. Results appear in the conversation panel and Analytics.

Analytics shows the verified CSAT average, survey response rate, 1–5 rating distribution, and a recovery queue linked to the affected conversations. The recovery queue always follows the low-score threshold selected by the workspace rather than using a fixed platform threshold.

Languages and replies

Language detection is cached with the conversation analysis. AI guidance is instructed to answer naturally in the customer’s detected language. A human can always modify the guidance before sending, and the approved result becomes part of the tenant’s learning history.