Effective Chatbots Should Meet Knowledge Management and Knowledge Centered Service (KM & KCS)

Effective Chatbots Should Meet Knowledge Management and Knowledge Centered Service (KM & KCS)

This article may seem odd for this publication, once it is not restricted to discuss only chatbots, intending to bring up some concepts about Knowledge Management and KCS — Knowledge Centered Service, and chatbots, of course.

First of all, it is important to stress out an important fact, that despite seeming to be obvious, sometimes it may go unnoticed.

The fact is, chatbots are not an end in themselves, they are used to achieve a goal.

That goal is invariably to be in touch with users or clients to inform, help to solve problems or to gather data. To do their tasks, chatbots must be properly trained. In other words we could say that without knowledge a chatbot will do absolutely nothing.

Before we go ahead discussing chatbots, let`s review some concepts about knowledge management — KM. According to the KM area, there are two types of knowledges, tacit and explicit. Tacit is everything that is not documented formally, whereas explicit knowledge , as opposed to tacit, is everything that is formally documented, stored, written, recorded, indeed, it is in another place besides people`s heads. Another important concept about KM is the knowledge conversion process, which means transforming tacit knowledge into explicit.

With that in mind, it is plausible to claim that training chatbots may be seen as knowledge transfer, as in some extent the bots must learn concepts and information that belongs to people or to organizations. This learning is achieved by using explicit knowledge, be it for understanding utterances or to give appropriate responses and information.

Knowledge, for companies, is an important asset. One of the goals of KM is to properly identify, collect, organize and disseminate knowledge. This is useful when we need to train chatbots, because we must know which assets (or knowledge) we have to transfer to our bots. This identification of knowledge is known as knowledge mapping.

When chatbots are trained and deployed, it is just part of the scenario. We all know that it requires a constant monitoring to review interactions with users and update the content, with a new cycle of training — as for new questions, new sets of information, new rules, etc. This is depicted in KM as proactive or reactive KM. The truth is, one way or another, one must monitor and update explicit knowledge.

Those who deal with KM know that is not easy or simple to keep explicit knowledge up to date. For this reason KCS was created. KCS is a methodology based upon practices which was created by a non-profit alliance of service and support organizations, called Consortium for Service Innovation. The fundamental precepts of KCS are creation of knowledge content as byproduct of the solution of the requirements, evolution of content based upon demand and use, development of knowledge base resulting from collective experience and collaboration and acknowledgement by learning, collaboration, sharing and improvement.

I do not intend to dive deeper in KCS concepts, but just to imply that some of the concepts could be very useful for chatbots training. Let`s take the first two precepts quoted above.

When users are interacting with chatbots, the undesirable situation is when a user asks a question and the bot is not able to answer — obviously it will not do it because was not properly trained.

There must be a process to update bot`s training when this happens, so why not take KCS`s precept of “ evolution of content based upon demand and use”?

This could means update bot`s training when it is not able to answer.


Some chatbots platforms support human hand-off, right? So, why not to take this opportunity to update the training when humans are escalated to dialogue with users? This is an “on-demand” knowledge update, as preconized by KCS.

Another example we could take from KCS is the “development of knowledge base resulting from collective experience and collaboration”.

When training our chatbots, who in the organization is assigned to do it? Would this person hold all the knowledge of all areas that is required?

So, why not collaborate on knowledge creation, and keep it cyclic?

Author`s note

For those who are interested in knowing more about knowledge management and KCS, I`d like to let a note about a book that I wrote, “Knowledge Management for Help-desk and Customer Care — How to build and effective knowledge base”, available at amazon.com.