In January, Facebook announced that it will be shutting down M, its text-based virtual assistant that used a combination of artificial intelligence and a few dozen human agents to interact with customers via Facebook Messenger. Where the AI technology fell short, the human agents stepped in and handled the more complex requests made by users.
M first launched in 2015 and never left beta mode. Only a few thousand people ever had access to it.
Despite the project's failure, M was an admirable idea by Facebook. M did what the more than 100,000 chatbots on Facebook's Messenger platform currently can not, like delivering a highly personalized, conversational virtual assistant that could handle any questions or requests sent its way. In the end, however, Facebook did not take into account the scope of the project they were taking on with M, which ultimately led to the virtual assistant's downfall.
It's impossible to ignore the chit-chat around chatbots. In fact, according to a report by Grand View Research, the global chatbot market is projected to hit $1.23 billion by 2025, an astronomical increase from $190.8 million in 2016.
But the majority of today's chatbots are backed by primitive AI technology and often can't do much more than answer simple pre-programmed FAQs. The interactions between chatbots and consumers are also confined to one platform, whether that's over Facebook, Slack, WhatsApp, or company websites, confining communication into one channel and creating a disjointed customer experience. What's worse is that many of these chatbots are promoted as AI when in fact they are often not, and when they don't work, they further suppress consumer trust around the technology.
Regardless of the hype, the concept of a chatbot as a standalone customer service solution is not realistic. As today's customers come to expect personalized, consistent, and intelligent customer care, companies must look to solutions that meet these rising standards.
What Facebook tried to do with M is exactly where the industry should be headed. By 2020, Gartner predicts customers will manage 85 percent of their relationships with enterprises without interacting with humans, demonstrating the need for companies to invest in flexible, reliable, and smart solutions.
However, when it comes to customer service, no task is inherently simple. Solutions must be able to understand the context of conversations and handle unexpected questions, which is exactly why Facebook looked to human agents to support M. But, due to the volume and complexity of user requests and the immaturity of generalized natural language processing (NLP), Facebook M never surpassed 30 percent automation and proved too costly to maintain.
How IVAs Do What M Could Not
While Facebook was unable to successfully marry the power of AI with the brilliance of humans, other technologies have completed the feat. These are called intelligent virtual assistants (IVAs). Using more than 10 years of data and training, as well as sophisticated automated speech recognition, NLP, and machine learning technologies, companies like Hyatthave successfully deployed IVAs to handle hotel reservations, saving millions of dollars, and ultimately providing a more positive customer experience. In contrast, Facebook's M had far less data at its disposal; more specifically, M had limited access to enterprise data and business rules and absolutely no previous history in customer care. On top of that, M still tried to take on the whole spectrum of user requests. Truth is, the interactions between business and consumers is extremely complex, something Facebook eventually learned with M's failure.
For most companies, chatbots are simply not enough. Eighty percent of communication between companies and their customers still happens over the phone, and the modern-day consumer demands consistency of experience across all channels, no matter how they decide to communicate. While chatbots are inherently text-based and limited to platforms like Messenger and web chat, IVAs are multimodal and can across all channels at the same time. For example, if a customer is making a hotel reservation with an IVA over the phone, that same IVA will soon be able to simultaneously show the customer images of their rooms on their laptops.
With Facebook M, the experiment showed that the combination of human and AI can do something truly game-changing for customer care. Getting this approach right is key to successful applications like Hyatt's. But the resulting solution needs to be made available on any channel, successful to any customer, anywhere.
Though the hype around chatbots is expected to continue, it's crucial for companies to understand the full spectrum of AI technology at their disposal. If companiesare looking to provide highly personalized, superior customer service and they find themselves asking whether to chatbot or not to chatbot, the answer is quite simple. Learn from Facebook and look beyond the chatbot.
Jane Price is senior vice president of marketing at Interactions.