Giving a voice to chatbots

by Cyril Lefebvre - Chatbot Technical Expert
| minute read

Although chatbots in the popular mind still tend to get associated mainly with written or text-based communication, spoken or voice-based communication in this particular application field of AI technology is clearly on the rise. At Sopra Steria, we are also looking into this promising new development. Especially since there is a real demand from some of our customers for this more advanced type of chatbot solution - if only for internal use.

I myself, in our Benelux DigiLab, have recently - and, I might add, successfully - been working on a proof of concept that links our own chatbot solution, Sopra Steria Alive Intelligence (formerly known as Botify.kit), to a voice recognition system, using a VoIP telephony service. One of the main objectives of my experiment was to extend the number of entry points for our chatbot platform, from plain old text or chat to, first and foremost, voice or telephony. I didn’t stop there, however, and also researched the possibilities of integrating with WhatsApp, Microsoft Teams, 3D avatars and other entry points that have become quite popular in recent years.

Simplicity is key

Another important goal behind my research experiment was to find a simple way of developing a contact centre environment based on AI-powered chatbots with voice recognition capabilities. These are the kind of chatbots that not only understand what a user is saying, but can also grasp his or her intentions within the context of the conversation. If necessary, they can even ask him or her for clarification.

The simplest way to reduce costs when developing such a contact centre, is through a platform that is easy to use. Ideally, that platform should allow clients to create and add their own call scripts or customer service scenarios at their own pace and liking. Using Alive Intelligence, our chatbot platform, our clients should be able to create all the chatbots that they need and want, link them to their other applications and allow their users to interact with them in a variety of different languages.

A couple of best practices

Relating to this need for simplicity, the best advice I can give to anyone thinking of creating their own chatbots is to start simple and go for the quick win. In other words, for your first implementation, think of the least complex solution that offers you the highest potential return on investment, whether in time, efficiency or money savings. Not only will you gain a greater acceptance from your colleagues, employees and customers by making a good first impression, but having shown the potential of the technology, you will actually end up doing more, not to mention more complex implementations with it.

With regard to voice-enabled chatbots, too, I specifically recommend that you keep their design as simple as possible. Although the technology for voice recognition has become quite good in recent years, there is always a margin for error. We only need to think of network issues that can cause a loss in sound quality and therefore in information. Advanced as our current voice recognition technology may be, in some cases it still pays to have a well-structured script or phone menu for the customer to navigate through. After all, the last thing anyone of us wants to do is frustrate and anger their customers.