• Aurenkin@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    15
    arrow-down
    7
    ·
    6 months ago

    You’d be surprised how often we can automate a customers enquiry with ML (not even generative AI). Humans are still there as a fallback, but it’s a way better experience to give instant help to the person if possible and then put them into the queue if they have a more complicated problem. Searching is not really in the same context as automating customer queries, although I guess it could depend on the domain to an extent.

    • stealth_cookies@lemmy.ca
      link
      fedilink
      English
      arrow-up
      2
      ·
      6 months ago

      If a request is simple and common enough that the request can be automated, then it is most likely something that I’m already pissed off about having to call in about since it should have been a feature on the company’s website.

      • VirtualOdour@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        3
        ·
        6 months ago

        Yes but how often do you call? People like us make up like 1% of their calls, 90% come from people asking ‘how do I download the Google?’ or ‘I saved a picture of my cat where did it go?’ and 9% is people with general questions they could have found online but they didn’t want to get into it.

      • Aurenkin@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        2
        ·
        6 months ago

        We use a home grown classification model for our customer facing stuff. There are some applications of LLMs we are using a SaaS for as it’s quick to get going but we are also working on fine tuning an open source model as well so we’ll see what ends up working better in terms of cost vs performance. That’s not going to be customer facing though, we don’t serve any generated text to customers.