Perhaps you are an executive considering a chatbot solution for your business, or maybe you are a super-forward looking marketer. Either way, in this post I am going to attempt to explain when chatbot technology is appropriate and if it is time to buy into the chatbot hype.
I do not want to go too much into the history of chatbots or dive into a narrative of the brands who were early adopters. I have written about this at length already.
Instead, I would like to talk more about the underlying technology and use cases powering the chatbot hype.
Over the last three months, the chatbots hype has started to quieten a little (in line with the Gartner Hype Cycle). Some had even started to get all super-doom-and-gloom, suggesting the end of chatbots in their entirety (FYI, chatbots are going nowhere - conversational software is the next paradigm in UX and human-computer interfaces). However, what we have seen is the sudden realisation that off-the-shelf and button-led chatbots are not very good, they are just glorified web forms. Therefore, naturally, people started to question: "what's the point of a chatbot?".
At the opposite end of the spectrum to off-the-shelf basic chatbots are those powered by natural language processing (NLP). When done well, these solutions are more simple and provide a more engaging and delightful experience for users (admittedly, when done badly they are worse then button-bots; hence Facebook's 180 in late 2016).
One of the earliest adopters of chatbot technology was Starbucks with their Virtual Barista. It was developed to enable users of their mobile app to order a drink via voice or typing. Lots of super-chatbot-hype people touted the Starbucks solution as the shining example of the benefits a chatbot can bring to an enterprise-level business.
With a critical non-hyped eye, the fact that this conversational solution was embedded into a mobile app is a bit cheat-y. After all, it meant the chatbot already had a massive user base of mobile app users along with jealousy-inducing million plus brand advocates who were willing to test new technology.
I think the Starbucks chatbot was good for the industry, but one of the many examples that led to the over inflation of expectations and hype.
Like all good business bloggers, I need to turn this case study into critical learning points for us to take forward.
There are two main requirements we can learn from Starbucks Virtual Barista:
You need an audience. If you are going to launch a chatbot via Facebook Messenger, then it is going to be tough going to learn and iterate with a small Facebook fan base.
You will need humans. A chatbot, even if it is a super-high-end one like one of ours, still needs to train the AI over time. Therefore, it will need people to take over the occasional conversation while it is learning. A chatbot does not always get it right; staff will be needed to deliver a seamless experience.
If your company does not have an audience or spare human capacity, then perhaps for you, chatbots may still be just hype.
There are some use cases where the chatbot hype is certainly not warranted. Chatbots are solving real business problems in sales, customer service and internal communication. I would recommend you start with one of these uses, the one where you have the greatest audience to test, learn and iterate quickly.
As an example, if your Twitter account gets bombarded with questions and customer service requests, a Twitter direct message chatbot to instantly answer questions and direct people to the right resource would be a smart addition to your tech stack.
Artificial intelligence (AI) is not an overnight solution or an instant cure to a problem. Also, they are not a siloed product that one person can manage. It takes some time to work with a development company to build and train the solution to your bespoke needs and, to succeed, will require deep integration with your systems and business processes.
However, some early adopters are proving the chatbot hype is real and is something that businesses of all size should be considering and testing.