Continuous refinement ensures your AI Agent consistently addresses key queries and delivers helpful responses, enhancing customer satisfaction. The Improve Bot section in Freddy AI Agent helps you refine your answers by analyzing live queries. It allows you to identify unanswered questions, review response correctness, and understand why some replies were marked unhelpful, ensuring your AI Agent delivers a better customer experience. 


The tabs in the Improve bot section include:

  1. Unanswered queries: Customer queries the AI Agent was not able to answer

  2. Answered queries: Customer queries the AI Agent answered. Admins can then review these responses to ensure they are accurate and appropriate. If an admin finds a response that could be improved, they can edit it or map the query to a more precise answer,

  3. Unhelpful responses - AI Agent responses that were marked as not helpful by customers.

You can also filter these tabs based on a timeframe by using the Filter button.

Analyze the following sections to improve your AI Agent's responses:

  1. Unanswered queries: Displays the queries for which the AI Agent could not respond. To improve AI Agent efficiency, you can either enhance your Knowledge Base with better content or add Q&As to train the AI Agent. 

Hover over an unanswered question to take any of the following actions:

  • Conversation history: View the conversation(s) in which the AI Agent could not answer the question.

  • New Question: Add this as a new query and map it to an answer.

  • Delete: Delete unanswered query.

  1. Answered queries: You can review the answers provided by the AI Agent. You should create Q&As to respond to certain queries with specific answers. Hover over an answered question to take any of the following actions:

  • Conversation history: View the conversation(s) in which the bot was able to answer the question.

  • New Question: Add this as a query and edit the answer or response to improve accuracy.

  • Delete: Delete answered query from the list of queries.

  1. Unhelpful responses: Check why some of the responses were marked as unhelpful by the customer and add additional information in your Knowledge Base or add Q&As.