Genesys Casual

 View Only

Sign Up

  • 1.  Copilot and Wrap ups codes

    Posted 5 hours ago

    Has anyone successfully implemented automatic wrap-up suggestions using Genesys Copilot in production?
    I would love to hear real use cases, lessons learned, and how accurate the suggestions have been for voice interactions.


    #CommunityQuestions(Contest,Community,etc.)

    ------------------------------
    Luis Antonio Padilla Yee
    na
    ------------------------------


  • 2.  RE: Copilot and Wrap ups codes

    Posted 4 hours ago

    Hello Luis.

    The most important thing when you want to work with wrap-up suggestions is to properly configure the code descriptions.

    Without this, you will probably face a lot of wrong recommendations.

    The key concept for Copilot for me is to reduce duplicated or similar objects and work with assertive objectives. 



    ------------------------------
    Arthur Pereira Reinoldes
    ------------------------------



  • 3.  RE: Copilot and Wrap ups codes

    Posted 57 minutes ago

    Hi @Luis Antonio Padilla Yee,

    Yes, we have used Copilot wrap-up suggestions in production, and one of the biggest lessons learned was that accuracy depends a lot on how clean and well-structured your wrap-up code list is.

    In our case, we had a scenario with many wrap-up codes, including similar, redundant, and sometimes obsolete options. When the list is too large and the descriptions are not clear, Copilot has a harder time distinguishing between similar classifications, especially in voice interactions where the suggestion depends heavily on the transcript quality and the clarity of the conversation.

    What helped us was applying a Pareto analysis to the wrap-up codes.

    We identified the tabulations that represented around 80% of the interactions and focused on improving those first. For these high-volume codes, we worked on clear, objective, and non-overlapping descriptions, making it easier for Copilot to understand when each wrap-up should be suggested.

    For the long-tail codes, we avoided overloading the model with too many similar descriptions at the beginning and treated them as a later optimization step.

    This approach helped increase the assertiveness of the suggestions because Copilot was no longer trying to choose between dozens of similar or poorly described options. Instead, it had a cleaner structure with better context for the most relevant contact reasons.

    So, my main recommendation would be:

    Before measuring Copilot accuracy, review the wrap-up structure itself.

    Remove obsolete codes, merge redundant ones when possible, and make sure the most frequent wrap-ups have very clear descriptions. In our experience, this makes a big difference in production adoption and agent trust.

    For voice specifically, I would also recommend validating the suggestions together with transcript quality, because poor transcription can directly affect the suggested wrap-up.



    ------------------------------
    Mateus Nunes
    CX Manager at Solve4ME
    mateus.nunes@solve4me.com.br
    Brazil
    ------------------------------