Why to DIY chatbots

Everyone Wants a Chatbot

One of the most surprising trends I have come across recently is the chatbot fad. Everyone is fighting to deploy some sort of chatbot as they dramatically improve customer experience. Chatbots scale much better than a workforce of live representatives which explains why seemingly every company wants to be on board.

The problem with chatbots lies in how companies are trying to develop them. By this I mean every other company seems to be big on planning on building a chatbot and debating over which platform to use but that is not affective as the platforms on the market cannot do what everyone wants.

What is wrong with the platforms on the market

Currently there are many different competing technologies such as Lex from Amazon, Watson from IBM, and Microsoft Bot Framework from Microsoft. The problem I have run into with all of these is they all have limits. Lex does not handle branching conversations well, Watson does branching well but is difficult to integrate deeply, and Microsoft Bot Framework has a high barrier to entry as it is so integrated.

The biggest problem with all of them however is you don’t know what you need until you are too far in. This happened two times into development of a chatbot with a company I was contracting for. We spent nearly a year and a half and had very little to show aside from some documentation.

We even started looking into buying a chatbot from another company but again ran into more problems than solutions as their chatbot was closed source so we could not build the integrations and the customizations we wanted with it.

Why to DIY

Stop talking about building a chatbot and build a DIY chatbot with the RASA stack. RASA is a python chatbot library that just works. Because everything is getting implemented using their open source library it has great transparency as to what is going on and how the chatbot is working.

RASA allows you to find out what you need from a chatbot quickly. If you need an integration, build it. If you want better analytics you can program it. After starting with RASA, I realized all the potential we were missing out on when we got locked into a platform.

Start with buiding a chatbot and you will find the how along the way.