Coaching knowledge bases

Coaching is an important aspect of knowledge management. After creating a knowledge base, you can start coaching to help the bot improve the quality of responses provided to the customers. With Zoom Virtual Agent, there are two levels of coaching that can be utilized:


After coaching is complete, you can begin training the knowledge base to ensure that the latest data from coaching is included and your ML model is updated.

Prerequisites for coaching knowledge bases

How to coach knowledge bases

  1. Sign in to the Zoom web portal.
  2. In the navigation menu, click AI Management and then Knowledge Base.
  3. To the right of the knowledge base you want to coach, click the ellipsis icon then select Coach.
    This will take you to the Coach tab where you can begin training queries.
  4.  In the Coach tab, you may search for existing queries or upload new queries to coach.
    • To coach existing queries, use the search bar to search for queries within the date range.
    • To coach a new query, click Upload Queries. Select or drag a CSV file with frequently asked queries from your customers to start coaching. You have the option to upload queries for a specific group, a particular knowledge base, or for all the knowledge bases you have.
  5. Enter keywords in the search bar to find the query, then select the query you want to coach.
    Suggested articles will be displayed as answers. 
  6. Select or create the answers that pair best with the selected query.
    • Select Article as answer: Allows you to search for or select suggested articles as answers to the query.
    • Create a custom answer: Allows you to select adjacent snippets in your knowledge base articles to provide more precise answers to your customers’ questions.
      Note: You can select up to 3 answers, which can be a combination of articles and spans.
  7. Click Coach.
    A message will appear, indicating that coaching changes have been made.
  8. Click Train.
    The progress bar will show that training and updating the ML model is in progress. A success message will appear once training is complete.