Chatbots are the hottest topics in the industry right now and are being looked upon as tools to completely transform marketing and other customer interactions.
According to a report by Forrester Research, globally, 57 percent of companies either use chatbots already, or plan to do so in the coming year. However, despite our best efforts, we have still not been able to create a perfect Bot. Facebook’s AI bot hit a 70% failure rate and Microsoft’s Tay completely failed the expectations and turned racist.
So Why do chatbots fail to deliver?
1) ChatBots created just because we can.
Most of the bots fail because of this ONE reason !
There is a mad rush towards building a Bot and organizations want to join this race and release a Bot as soon as possible. In all this craziness, they fail to properly analyze if they even need a Bot for a particular purpose. This is common with every new technology. Developers and designers get over excited about the new technology (example Mobile App in recent years) and end up churning a number of irrelevant use cases.
Lately a lot of chatbots were created by companies to replace their automated voice responses. This means, the functionality of the bots was exactly the same as an automated response system, with a set of simple, static responses. Bots such as these, are doomed to fail.
It is thus very important to understand the necessity to create a Bot, and then define the right plan for it.
So before you go ahead with deciding the deadline for the release of the next chatbot, take a step back, define a use case, analyse the business justification, discuss the scope and then start the implementation !
2) Chatbots created in silos
Another very common reason for failures of chatbots are when companies try to create them as completely new functionalities. Bots should ideally be part of the existing systems and should be able to communicate with them. Example : an organization created a bot for customer support queries. But there already exists a Customer Support Call Centre for the same. If the chatbot does not communicate with the existing channel in some way, it would just tend to become an overhead for the business, as well as a frustrating experience for customers.
3) Chatbots created for doing “everything”!
The best examples of this concept are “Siri and “Alexa””. These bots were not built for any One purpose. For them, to understand “everything” is impossible. The new Bot in the same domain called “Viv” also faces the same difficulties. Humans learn over their entire lives. So to correctly imitate the same behavior and responses, is extremely difficult. Add to it the irrational aspects of human behavior, and you have an even bigger problem in hand.
For the above mentioned reasons, we came across scenarios like these :
It is always a good idea to limit the scope And give enough time to the Bot to train and understand the different aspects of human conversation.
4) Chatbots cannot be “Human”
Human conversations are complex. We understand emotions, accents, tones and always rely on a lot of contextual information, when giving a response. Even with NLP and AI, these aspects are immensely complex to imitate and might take years of training to perfect, if at all.
The moment customers start feeling that the responses are lacking the human touch, and perhaps are out of context, they get frustrated. Bots like Alexa are expected to carry an intelligent conversation. Although the users are completely aware that they are communicating with a machine, they still get impatient with irrelevant responses. The only way to ensure this does not happen, is to spend time in training the Bot.
5) Perfect Chatbots need better technology
Many companies expected bots to take over all the human interactions with their customers. Plans like these are a perfect recipe of failure. Success of Bots should be defined as some Value Add, and not something unattainable.
The technology giant Facebook announced their M project in 2015 and as of 2017, there are around 10,000 users. Facebook’s M project, as of now, can only respond to 30 percent of customer requests. M is still learning and getting better. If M does not understand the user request, a human takes over and that is how M learns. So if we are envisioning a perfect chatbot, we need a better and more mature technology.
