An emerging technology for call center is the usage of voicebots. The main purpose of introducing bots is to reduce the load from the call center agents. Voicebots are usually the first to receive the telephone call from the customer, provide initial assistance and if required route the call to the best human agent group based on their skills. The bots are analyzing the speech of the customer in order to understand the customer’s request.

In order for voicebots to work efficiently, they need to be able and analyze correctly the spoken words said by the customer. If the voicebot is unable to accurately understand the spoken words, it will end up forwarding the call to a human agent without doing any work/filtering. As a result no time saving will be done and on the contrary the customer will be upset since his/her time was wasted on a void discussion with a bot.

There are few factors that dramatically impact the accuracy of the bots in understanding the customer’s speech like: accent, vocabulary and noise. In this post we will discuss how to address the latter factor, the ambient noise. Customers can dial into the call center from almost any environment. As a result, there might be phone calls with significant ambient noise like street or airport noise. The call center has no control over the ambient noise but it should try and cancel it before it reaches the bot since a significant amount of ambient noise has a good chance to dramatically impact the performance of the bot.

In order to cancel the ambient noise that is coming from the customer side, the noise-cancellation filter needs to be placed between the customer and the voicebot. This means that the noise cancellation needs to be done in the VoIP network itself. A typical way of doing so is to run the noise-cancellation as a SIP Trunk since Sip Trunk is the most common plug-and-play technology for VoIP networks. It is important to note that if you place the noise cancellation on the agent’s PC, it will not be able to improve the accuracy of the voicebot since in this location it removes the noise coming from the customer side after it was already analyzed by the voicebot.