Okay, let’s start by getting the compulsory buzzwords out of the way, shall we? Artificial intelligence, natural language processing, machine learning. There. It seems that any discussion of almost anything tech-related nowadays has to include at least two of the three. We’ll return to them later. So, chatbots. Those rather annoying automated messaging algorithms that you’re likely to encounter in lieu of a real, flesh and blood, customer service agent. Are they just a flash in the pan? A solution looking for a problem? Or are they on the cusp of revolutionizing how we do just about everything, from interacting with the bank to shopping for flights. Cue obligatory bullish statistic: According to a survey of 800 decision-makers conducted by Oracle, 80% of businesses are expecting and/or preparing to be using chatbots by 2020.
A Very Brief History
It’s interesting to note that the idea of chatbots goes all the way back to the very dawn of Artificial Intelligence (buzzword number 1) research. In 1950, mathematician and computer scientist Alan Turning laid down the gauntlet when he proposed perhaps the most famous test of machine intelligence. Human evaluators would blindly judge text-only conversations between real human beings and computer programs, trying to determine which was which. If the program could convince the evaluator that it was a human being, it was said to have passed the Turing test.
The first ever chatbot was actually an attempt to take up Turing’s challenge and provide a critique of its criteria. Developed by Joseph Weizenbaum at MIT in 1966, ELIZA was programmed with almost zero information about human emotion or thought and yet was still able to interact with human beings in an engaging manner, even when the human went beyond ELIZA’s limited knowledge base. ELIZA was modeled on the psychologist Carl Rogers’ technique of eliciting information from his patients through the use of open-ended questions and demonstrated that strikingly realistic interactions could be had by using keyword search and a large database of pre-prepared responses.
General Vs Narrow, Strong Vs Weak
50 years of technological improvement have taken place since then. Siri and Alexa are certainly light years ahead of ELIZA in terms of their natural language processing (buzzword number 2) capabilities, which involve speech recognition, understanding and generation. However, advanced as they are, they still work along fundamentally similar lines, relying on a search function paired with an enormous database. Even IBM’s Watson, who in 2011 beat the human Jeopardy! champions, employs this same brute force approach. Many of the largest leaps in the field have occurred in narrow or “weak” AI, with general or “strong” AI seeming further away than AI researchers initially thought.
However, these narrower applications are perfect for businesses keen on categorizing and automating the interactions they have with their customers. The automated phone system of your utility provider is like a 1.0 version of this, but the idea truly gets its wings in the smartphone domain. Consider that the top three messaging applications in the world, WhatsApp, FaceBook Messenger and WeChat boast around 2.5 billion users between them. This amounts to a captive audience that can be encouraged to use the same chat apps they know and love as a first point of contact, satisfying customers and cutting customer service expenditures in the process. According to Juniper Research, businesses could potentially save up to £6 billion per year by deploying their own chatbots, with banks currently saving an average of $0.60 per chatbot interaction. A 2017 poll by HubSpot Inc revealed that 40% of consumers don’t care about whether a chatbot or human assists them. It’s safe to say that this number is likely to rise as they improve and become more commonplace.
It’s Just a Matter of Time
Customer support is a no brainer as far as chatbots are concerned, particularly as an initial point of contact for more routine communications, which can then be escalated to a live human agent as and when the need arises. But marketing and sales are truly where AI chat bots are set to shine. From informing users about products and services to actually selling products directly, the growing trend is that chat is starting to eat the world. According to Apptopia, out of the top five most used apps of 2018, three were chat apps (WhatsApp, WeChat, Facebook Messenger). We are spending a lot more time on our chat apps and businesses are starting to take notice.
None of this comes as a surprise to Chinese smartphone users, who have long been accustomed to using WeChat for everything from sending each other money and paying bills to booking doctors appointments and tickets. We’re slightly lagging behind in the West but the popularity of Internet banks and the explosion in digital currencies are signs that the financial and digital realms are finally converging. It’s just a matter of being able to do everything from the same interface. Imagine receiving a tailored special offer from your favorite online retailer via chat and being able to complete the transaction simply by telling the bot what you want. It’s not as far off as it may sound. The possibilities are endless, especially when you factor in referrals and in-content advertising. Chat platform Kik has been experimenting with influencer bots that pop up on behalf of your favourite social media stars and offer you coupon codes and the like.
And They’re Only Getting Smarter
This technology truly comes into its own when you factor in machine learning (buzzword number 3). If you think about it, chat app users are an untapped resource of enormous amounts of data. Consider the leaps forward in terms of market research and product development that chatbots can offer, especially when they’re performing everything from customer service and marketing to sales for us. Now, imagine data from each of those domains being fed into the others in order to make progressively smarter bots. Even relatively simple market research bots like Swelly, which collects data by getting users to pick between two choices on social media, offering to reveal what others have selected as a hook, are able to garner a tremendous amount of actionable data that can then be used to perform more effective online marketing.
Bots can also be multiplied in ways that human employees simply cannot. As they become smarter and more capable they’re going to be able to provide customers with personalized experiences at scale. This is pretty much the holy grail for all businesses. Everyone pays lip service to it, the words “tailored” and “bespoke” get thrown around a lot, but hardly anyone really does it right. As chat bots become ubiquitous even small businesses will be able to offer this to their customers, which could also lead to a massive leveling of the playing field.
Watch This Space
In 2016 the value of the chatbot market was estimated by Markets and Markets Research at $703mln. According to Opus Research, $4.5bln is expected to be invested in chatbots by 2021. We’re talking massive potential for growth here, and in an area that doesn’t even need to worry about building its own platforms and attracting users. This ability of chatbots to piggyback on already established platforms with enormous userbases makes them potential game-changers. They may just be a recent trend, but they definitely have a bright future ahead of them. Let’s just hope they stop at serving us. It would be tragically comical for our future machine overlords to start life out as humble chatbots.