Call transcriptions

STRUCTURE OF A TRANSCRIPT

Each call transcript consists of a series of timestamped messages, identified by their persona:

• assistant: Responses generated by your AI bot
• human: The caller's words, automatically transcribed
• tool: the bot's internal actions (knowledge base search, etc.). These technical messages are part of the history but are not shown in the transcript view, to make it easier to read.

AUTOMATIC SUMMARY

At the end of each call, our AI automatically generates two types of summary:

Short summary: A one- to two-sentence recap that captures the essence of the conversation. Ideal for a quick overview in the calls list.

Long summary: A detailed summary that covers the key points of the conversation, the client's requests, the responses given, and the actions to take.

SATISFACTION LEVEL

For each call, the AI estimates a satisfaction level from five categories (Very dissatisfied, Dissatisfied, Partially satisfied, Satisfied, Unknown), based on the caller's tone, the resolution of their request, the fluency of the exchange, and detected expressions of satisfaction or dissatisfaction.

This level helps you quickly identify problematic calls and improve the quality of your AI customer service.

FINDING A CONVERSATION

To find a specific conversation, use the search bar at the top of the calls list (by caller number or email), as well as the channel and date filters. Then open the call to read its full transcript.