Genesys Cloud - Main

 View Only

Sign Up

  • 1.  Designing an AVA for an older audience: 82% retention in the first weeks

    Posted 9 hours ago

    We recently launched a voice Agentic Virtual Agent for a financial institution, focused on supporting a customer base with a strong presence of older adults.

    The use case started with a very common behavior among these customers. Many call frequently, sometimes several times throughout the same day, to check their account balance, confirm the amount of a benefit payment, or verify whether the credit has already been deposited.

    Although these requests may seem simple, they represent a significant volume of repetitive calls. At the same time, this audience requires an experience that is clear, patient, accessible, and flexible.

    Another important characteristic is that customers do not always express their needs through a direct or structured request. They may begin by explaining what happened, sharing personal context, or telling a longer story before finally asking about their balance or benefit.

    This conversational behavior was one of the reasons an AVA was a strong fit for the project. Instead of forcing customers to navigate a rigid menu or use a specific command, the AVA can follow the conversation, interpret the context, and identify the actual need, even when the request only becomes clear after a longer explanation.

    Combining transactional self-service and knowledge

    The AVA was designed to provide:

    • Account balance information
    • Benefit payment information
    • Both pieces of information during the same interaction
    • Answers to common questions about the bank's mobile app
    • Guidance on downloading and accessing the app
    • Instructions for completing transactions
    • Support for other common digital banking questions through a knowledge base

    The knowledge component is especially important because many customers call not only to retrieve financial information, but also because they need assistance using digital channels. By combining transactional self-service with knowledge-based support, the AVA can resolve the immediate request while also helping customers become more comfortable with the bank's digital services.

    Initial results

    From July 22 to August 5, the AVA handled 325 calls, with the following results:

    • 82% retention: 268 calls were completed without reaching a human agent
    • 78% self-service usage: 254 customers received their balance, benefit information, or both
    • 70% complete resolution: 226 calls were fully resolved after the requested information was delivered
    • 62% of calls included a balance inquiry, representing 200 interactions
    • 45% included a benefit inquiry, representing 146 interactions
    • Only 17.5% of calls were transferred to a human agent
    • PSAT of 4.0 out of 5, with 75% of respondents rating the experience with a 4 or 5

    A single call may include both a balance and a benefit inquiry, which is why those percentages overlap.

    The PSAT result is currently based on 32 responses, with 24 customers providing a rating of 4 or 5. We are therefore treating it as a positive initial signal rather than a consolidated indicator and will continue monitoring it as adoption grows.

    A controlled rollout before expansion

    The current volume is intentionally limited because the AVA is initially available only to customers from selected area codes.

    We chose a phased rollout so that we could monitor the conversations closely, understand how this audience interacts with the AVA, identify opportunities for improvement, and validate the overall experience before expanding it to a broader customer base.

    Beyond the numbers, we also received very positive feedback from the customer team. They highlighted the quality of the delivery, the collaboration between the teams, and our availability to support them whenever needed.

    For me, this project reinforces an important point: designing automation for an older audience requires much more than technical accuracy. Language, pacing, clarity, patience, repetition, error recovery, and the ability to understand less structured conversations can determine whether customers feel confident enough to continue the interaction.

    These first results show that voice automation can reduce repetitive demand, provide immediate access to important information, and support digital inclusion while allowing customers to communicate naturally and in their own way.


    #ConversationalAI(Bots,VirtualAgent,etc.)

    ------------------------------
    Mateus Nunes
    CX Manager at Solve4me Solucoes em Tecnologia Ltda
    ------------------------------


  • 2.  RE: Designing an AVA for an older audience: 82% retention in the first weeks

    Posted 9 hours ago

    Matheus,

    First of all, congratulations! AVAs are still a relatively new technology, and it's impressive to see how your team is discovering new ways to apply LAM models to improve the customer experience.

    What stood out to me the most was the focus on the end user. Designing a conversational experience for an older audience requires much more than technical accuracy, it requires empathy, patience, and the ability to understand conversations that are often unstructured. Achieving an 82% retention rate and a 70% resolution rate in the early stages of the rollout is a great result.

    I also really liked the decision to start with a controlled rollout. Monitoring real conversations, learning from customer behavior, and iterating before scaling is exactly what helps build a reliable and effective AVA.

    One thing I'm curious about: during the rollout, what turned out to be the biggest challenge in helping the AVA understand those less structured conversations from older customers? I imagine there were some interesting learnings along the way.

    Congratulations to everyone involved. I'm looking forward to seeing how the solution evolves as it reaches a broader audience.



    ------------------------------
    Lucas Santana
    Pre-Sales Consultant at Solve4ME
    ------------------------------



  • 3.  RE: Designing an AVA for an older audience: 82% retention in the first weeks

    Posted 6 hours ago

    Thank you so much for the thoughtful feedback!

    The biggest challenge was exactly the lack of structure in many of the conversations. Customers often explained the situation through longer stories, provided information out of order, changed the subject midway, or used expressions that did not match the terminology configured in the journey.

    Our main learning was that improving the AVA was not only about adding more examples or instructions. We also had to make it more tolerant of incomplete information, maintain context across the conversation, and avoid responding too quickly before fully understanding the customer's need.

    Monitoring real interactions during the controlled rollout was essential. It allowed us to identify recurring communication patterns, adjust the guidelines and knowledge content, and make the experience more patient and natural without compromising accuracy.

    There is still a lot to learn, but these early results have shown us how important it is to design the AVA around the way customers actually communicate, rather than expecting customers to adapt to the technology.

    Thanks again for the question and for recognizing the work of everyone involved!



    ------------------------------
    Mateus Nunes
    CX Manager at Solve4me Solucoes em Tecnologia Ltda
    ------------------------------