There have been quite a few advancements around how AI voicebots have improved the accuracy and help call centres out, providing a much better level of accuracy. With greater accuracy, this leads to a better overall experience for those call centre businesses and the customers as well. Yet it hasn’t been really at a perfect level, with only a 65% success rate for major vendors. So how is this changing?
It’s important to note that regardless of which direction the call is happening, a voicebot needs training and help to understand that it's not talking with a human and is, in fact, speaking with a voicemail. Regardless of what the next message is, that voicemail detection is critical. Otherwise, the voicebot and voicemail will talk at the same time.
It all starts with typical voicemail language. It sounds like a real conversation and begins with a lot of rambling. Then there's the fact that there's no 'standard' or importance of getting the voicemail right. Its purpose is not to work with voicebots but to give the caller a chance to leave a message for a future follow-up.
Even if a voicebot is smart enough to understand that it’s speaking with a voicemail, it will wait, but too many gaps between the recordings will have it think the voicemail is done before, and that means the voicebot will start to activate.
So now that we see some of the challenges that call centres can face, instead of simply throwing away those calls that go straight to voicemail, getting the right type of voicebot trained is of utmost importance.
Its predecessor, the auto dialler, have components when it comes to understanding different telco messages or different formats of an answered response. These diallers have been the main tool up until now and they are pretty dumb when understanding that they are connected to a machine as well. The industry standard of accuracy for these is as little as 65%, but there is no need to lose out on one-third of these calls just because they go to voicemail.
Voicebots have the advantage of using powerful NLU (natural language understanding) that helps them to be that much smarter with the voice it is listening to. However, not only is understanding the language important, but it's also important to factor in the gap in the messages and not just start chatting right after the first pause.
We have the technology and the capacity now to build out a wide enough range to be able to incorporate all the variations of voicemails, and instances, so that a voicebot has now raised the accuracy into 90%+ without further tuning, truly understanding through its learning that it is talking toa machine as well.
For those that see the need and are ready to take the next step in the evolution of voicebots looking to overcome the issue with the myriad of variations of voicemails and lets not forget IVR answering calls, please feel free to reach out to us to consider solutions.
We have our podcast about a variety of subjects including chatbots, voicebots as well as customer service or self service automation, we discuss how these impact business and some specific challenges or sector stories to help anyone looking to engage with this technology.