Genesys Cloud - Main

 View Only

Sign Up

  • 1.  Suggestions for detecting when Messaging contact is a GenAI bot/agent and not a live person?

    Posted 5 days ago

    Good day,

    In the past, we have had bad actors try to flood our queues using bots that could follow our IVR or respond to the standard messages in our SMS and WM flows. We had developed some best practices to try to eliminate them without impacting people needed to connect our advocates in crisis situations. However, we have seen an uptick in what we call "hangups" which are interactions that wait in queue and then when our advocates answer, either they don't respond or they respond with something completed unrelated to the services we provide.

    I'm wondering if anyone has come up with a way to detect this situation without adding too much friction to the interaction initiation process?

    Would love to hear people's thoughts.

    Thanks


    #DigitalChannels

    ------------------------------
    Marty Hand
    CTO
    ------------------------------


  • 2.  RE: Suggestions for detecting when Messaging contact is a GenAI bot/agent and not a live person?
    Best Answer

    Posted 5 days ago

    Hi Marty

    I haven't been through that myself, but I thought of something.

    Could you implement something similar to CAPTCHA... 

    For example, asking the user to reply with a specific word before being connected to an agent (e.g., "Please reply with HELP so we can connect you with an advocate"). Simple bots and queue-flooding scripts often fail these checks or ignore them entirely.

    Automated systems often respond with generic or unrelated answers, which can help identify suspicious interactions early and includ the decision block in the flow, if the response is the same as requested (e.g., HELP), it transfers, if not, it disconnects.



    ------------------------------
    Kaio Oliveira
    GCP - GCQM - GCS - GCA - GCD - GCO - GPE & GPR - GCWM

    PS.: I apologize if there are any mistakes in my English; my primary language is Portuguese-Br.
    ------------------------------

    Original Message


  • 3.  RE: Suggestions for detecting when Messaging contact is a GenAI bot/agent and not a live person?

    Posted 5 days ago
    Edited by Phaneendra Avatapalli 5 days ago

    Hi Marty,

    That's a good suggestion from @Kaio Oliveira, especially as a low-friction first step.

    You could also build on that with something like a button or carousel-style check before queueing, for example asking the user to choose the correct option before entering the bot or being transferred to an agent. This can work alongside the keyword approach, as some basic bots may find structured interactive responses harder than simple text replies.

    If you do go down the button route, another advantage is that a genuine button click can send a postback payload rather than just plain text, which could be used as an additional validation step in the flow.

    I had actually designed something similar using a carousel for a callback bot, although we never ended up implementing it as we have not seen security concerns on that bot so far.

    The key would be to keep it simple so it doesn't create too much friction for genuine users.

    Hope this helps and keen to hear how others implemented.

    ------------------------------
    Phaneendra
    Technical Solutions Consultant
    ------------------------------

    Original Message


Related Content