Introducing the “Customer Behaviour Understanding” -CBU- box in the chatbot architecture.

Since we started we were asked often…”but are you creating chatbots?” “No? then you are working in natural language AI, aren’t you?” Well, we are not creating chatbots nor working in the needed language tools, we are trying to create a new category of product adding value not only to the chatbot ecosystem but the customer care as a whole. We are sure that chatbots will become a mainstream customer to enterprise communication but adding useful info from different pre-existing channels can improve the chatbots.

In order to introduce this new box it worths to see the state of art, which are the current technologies needed and used to create a chatbot. This post by Anush Fernandes (from Verloop) explains very well the current chatbot architecture and their “boxes”.

diagram by Verloop.io

According to the diagram the artificial intelligence, machine learning is focused in the understanding the language (NLP/NLU) and the decision engine.

In Anbotux we believe that understanding the language is only a part of understanding the user, the customer. Each customer has an story behind and beyond the chatbot interaction. The chatbot in the sense of “conversational” can provide a lot of insights about the customer in the same way that voice and the direct customer facing person in a physical office. 3 decades ago with no mobile apps, web and even no or just little phone customer care the customers of a given enterprise were tracked and helped by persons that understood them, knew about the humour, concerns, risk tolerance, churn probability. The aim of Anbotux is rebuild that type of customer base-knowledge combining the story of interactions in all customer care channels, the segmentation data from the enterprise, the external contexts that can impact in customer actions and of course, the behavioral insights that chatbot interaction will provide. It is the goal of this new box “Customer Behaviour Understanding (CBU)” and where is located in the chatbot architecture diagram:

Then, artificial intelligence and machine learning techniques may be applied in another level and reusing some datasets derivated from information that current intensive in customer-care enterpriese already have and also using external phenomena contexts. Discovering patterns for a given customer or for a given particular segment of the customer base will enable a real-time chatbot personalization in two dimensions:

• Personalization the interaction with relevant content according to the customer learnt concerns and/or preferences.
• Personalization the interaction adapting the chatbot style to the way that the customer interacts better (shorter o larger messsages, menu preference, etc).

Last but not least important is how this new box is integrated with the other pre-existing boxes. Painless integration is key and the approch is API based, both ways: API to load the data from each data source and API too to send the personalization info to the chatbot developer in charge of the “Decision Engine”

Whatever feedback, suggestion, concern you leave in the comment would help to improve this post :-)

Follow us in Twitter @anbotux