How to succeed with Chatbots
Chatbots are still one of the biggest hypes, even though so many chatbots have failed. Do chatbots deserve to be a hype, Yes definitely, we are still just in the start of this era.
Why have so many chatbots failed and how should you do to avoid the failures?
First of all, let us define a chatbot. Chatbots are a service powered by artificial intelligence that you interact with via a chat interface, an alternative to human conversation. Chatbots are good at specific tasks, simple ones like a rule-based chatbot that answers basic customer service questions, or more complex operations, like helping customers find the right products based on their style and needs. Two great examples of chatbots that works fine are Booking.com’s Messenger chatbot and Finnair's ”Finn” Messenger chatbot. Chatbots work great when integrated with apps and websites, in order to boost interaction with customers, because of a change in user preference towards conversational marketing. They are not a be-all, end-all solution for your company, but if they’re used right and made properly, they can be a consistent, context-driven, conversational and cost effective way to change the way your business operates.
So, why have so many chatbots failed ? The chatbot field is one of the most complex area of Artificial Intelligence (AI), so obviously creating a chatbot and not mastering AI will not yield a good result. There are, however, other aspects as well, so let us look at reasons, why chatbots fail:
- Poor chatbot dialogue design
- The chatbot hype, in itself
- Wrong chatbot tool
- Poor Artificial Intelligence
Poor Chatbot Dialogue DesignVery few people that are responsible for designing the customer experience have experience with chatbot technology. To allow your IT department to own the chatbot project increases the chances for you getting a poor chatbot solution, even further. Are your IT department, your customer domain experts? You should have some experienced Chatbot UX designers to lead the project.
The Chatbot hype in itselfMedia of all types and every analyst firm and consulting firm tells you, you got to have chatbots. Yes, you probably do. At least if customer experience is your competitive battle ground. And with the proliferation of mobile devices, you need chatbots as an integral part of your digital transformation. However all of this hype about chatbots has a slew of negative consequences. You end up experiencing pressure from all over, and you might be pressed into a Chatbot-project, where you need to make decisions in a haste, which leads to a lot of negative consequences as well, none of them good. This is something everyone experience.
Wrong Chatbot toolToday, there are several generations of chatbots, each of them can deliver business value. Scalability is a common problem as well as functionality. A 1st generation chatbot is a very simple rule-based engine. You have very common questions, you can program very simple answers. These bots are easy to build easily bought and easy to fail with as soon as your customer expects more, as they always do. The 2nd generation chatbot leverages supervised machine learning as a complement to the rule based concept. Finally the 3rd generation chatbots can even do unsupervised machine learning. Spend some time in finding out what your business objectives the chatbot is intended to support, this will give you some guidance in choosing the appropriate chatbot tool.
Poor Artificial IntelligenceAs, mentioned earlier, the chatbot field is one of the most complex area of Artificial Intelligence (AI), hence good tools and a good understanding of AI is needed to create a successful chatbot. One of the fundamentals in an AI driven chatbot is semantic text analysis, ie, the computer's ability to understand what a human is trying to tell the machine. A majority of the few successful chatbots are in English or Spanish. I have still to see an advanced successful chatbot managing a good dialogue in Norwegian, Swedish, Finnish or Danish. Hence, one of your requirements should be your language. Has the tool you are looking at, a language package in the intended language for the chatbot? Do the vendor/consultancy company that offers you a chatbot have any documented experience in NLP (Natural Language Processing), NLU (Natural Language Understanding) and NLG (Natural Language Generation)?
Ok, so I have found my text analysis expert and a tool package in my language, then what? Well, Artificial Intelligence needs to learn. You can dramatically decrease your failure when deploying your chatbot by giving it a good head start. You have to train it, feed it data, and nurture it. The time it takes for your chatbot to learn can take a few weeks to a few months. This variance is somewhat impacted by volume but more importantly it is impacted by the quality of your chatbot solution. Is, for example machine learning applied to your dialogue setup? Machine learning will have a huge impact in learning and ongoing nurturing and maintenance of or your chatbot. One, what I would call a simple language solution feature, that I often see missing, is a synonym list in the language package. Clearly, one has forgotten this, in Finn (Finnair’s chatbot).
In my dialogue with Finn, I want to travel from Oslo to Peking (the Swedish naming of Beijing). Finn does not recognize Peking, but as soon as I enter Beijing Finn understands. With an implemented synonym wordlist, stating Peking=Beijing, the bot would have understood at once.