If you follow the digital marketing field, you probably have lately run into headlines like “Chatbots are taking over the world” or “Why you should build a chatbot”. Many brands are developing their own Facebook and Twitter bots and conversational user interface is becoming more and more popular on apps and websites. Bots are also taking the role of customer service, something they can do 24/7.
Chatbot and conversational UI are definitely buzzwords of the moment – but are they more than that?
Oldest trick in the book
Chatbots are actually older than internet itself – the first notable chatbot, Eliza, was introduced in 1966. Despite being a very simple language parser it was, at the time, actually able to convince some users that they were talking to another human.
Chatbots have been there throughout the history of internet. They were an essential part of IRC in the 90s, as well as the first online chat rooms a bit later. With smartphones came the virtual assistants, Apple introducing Siri in 2010.
So why are chatbots in such a hype right now? Maybe because platforms like Facebook Messenger and WhatsApp have made the use of conversational UI feel natural and easy. Introducing chatbots to masses as customer service or marketing tools is an easy step. We’re chatting all the time anyway.
So, is this AI?
Most chatbots still basically work the same way as old Eliza: they take the user’s input and react to certain words / sentences and then generate the response according to that. Sometimes chatbots give pre-defined options – those are essentially just web forms in new, fancy clothes. (There’s even a framework that turns an ordinary web form into a conversational UI.)
When used inside a social media platform like Facebook, a chatbot can also gain access to users’ (public) account data. By using the user’s name and generating different responses based on their age, gender or location, the chatbot may seem to really know a lot about you. But there’s still no “real” AI behind it.
There are, however, bots that use actual machine learning. The most notable ones being Siri and Alexa, who, by gathering data about their user, become able to give better and better responses over time. The more infamous example would be Microsoft’s Twitter bot Tay, who turned into a racist troll in less than a day. In a way the bot worked as it was intended: it learned from other users and started to behave just like them.
However, in overhyped articles about chatbots, terms like machine learning and artificial intelligence are thrown around quite lightly. In reality, most chatbots have no complex AI behind them. And for purposes they are most used, they don’t really need one.
“Should we have a chatbot?”
Short answer: not just because. Just as with all overhyped concepts, many companies will introduce their version just to ride that wave. Because everyone else is. This will only result in an armada of shitty bots that serve no real purpose. The same thing happened when native mobile apps became popular – everybody had to have one. Most got 0 downloads.
If, however, you see a real use case for one – then go for it! Ask yourself if the chatbot would make the user experience better, more enjoyable or smoother. Do you have actual content best served in a form of a conversation? If the answer is yes, the tools for building a cool chatbot are now better than ever. Let’s do it.