Watch the video presentation above or read the full article below.
Increasingly, whether it’s on Facebook Messenger, Slack, Kik, or WhatsApp (which combined have a user-base of more than one billion), we are interacting with chatbots. Bots are performing a variety of tasks for us, from weather updates to traffic alerts. They are even holding full conversations! Facebook and many other tech giants, including our partners at HubSpot, are betting big that within a year or two, bots will be responsible for many of our customer service interactions and will change how we buy online. According to Microsoft CEO, Satya Nadella, “People-to-people conversations, people-to-digital assistants, people-to-bots…that’s the world you’re going to see in the years to come.” In fact, Gartner predicts that by 2020 “the average person will have more conversations with bots than with their spouse.”
A chatbot is defined as a program designed to mimic human-to-human interaction via text or voice inputs. For the purposes of this post, our focus will be on text bots, such as the tens of thousands that have been built on Facebook’s Messenger platform since it launched in April 2016. Digital assistants that are controlled by voice such as Apple’s Siri, Microsoft’s Cortana, and Google Now are a more evolved form of chatbot. We will cover these and other bots that use natural language user interfaces (UI), or skills, in our next presentation in the 2020: Future of Marketing series, “Artificial Intelligence (AI) & Machine Learning.”
Chatbots have been called both the “new internet” and the “new apps.” It may seem hard to believe that such a relatively simple program could have such a radical impact on how your audience will interact with your organization. However, the bot revolution is well underway and as Purna Virji, Senior Manager of Global Engagement at Microsoft, recently remarked during her presentation at Inbound 2017 on chatbots, “caution is necessary, avoidance is not.”
A Brief History of Chatbots
Experiments with chatbots date all the way back to the 1960s when MIT professor, Joseph Weizenbaum, developed a bot named ELIZA, after the character Eliza Doolittle from Pygmalion. Weizenbaum initially built ELIZA to parody “the responses of a non-directional psychotherapist in an initial psychiatric interview.” If you were to tell ELIZA that your head hurts, ELIZA’s response would be, “Why do you say your head hurts?” Weizenbaum actually intended to prove that interactions between humans and computers would only ever work on a superficial level, but he surprised himself when he found many who interacted with ELIZA began to form a recognizable emotional bond.
The next major chatbot advance came in 1971 with Parry, a natural language program that was developed to simulate the responses a paranoid individual would give. Parry was programmed to operate under the assumption that “others must be up to no good.” As odd as this program sounds, it was the first chatbot to pass the Turing Test – a test developed by Alan Turing in 1950 to test a machine’s ability to become indistinguishable from a human. Subjects interacting with Parry and actual human responders were unable to distinguish, with more than random accuracy, Parry’s responses from those of a paranoid human.
Chatbots, such as Jabberwacky (1988), Dr. Sbaitso (1992), and A.L.I.C.E. (1995) continued to evolve their ability to mimic human responses and understand natural language, but the next major advance came with SmarterChild (2001)*. SmarterChild’s contribution to the evolution of chatbots was its ability to pull and return information from the then-nascent Internet. If you ever used America Online Instant Messenger or AIM then you might be familiar with SmarterChild. Residing within your buddy list, Smarterchild could answer your queries on stocks, movie times, and weather.
Another of SmarterChild’s innovations, one that can be argued as missing from many of today’s AI digital assistants, was a distinct personality. As described herein “Motherboard’s A History of Smarterchild”:
The Beginning of a New Internet
“Chat apps will come to be thought of as the new browsers; bots will be the new websites. This is the beginning of a new internet.”
Ted Livingston, Kik
In a widely circulated blog post from 2016, Livingston challenged the prevailing idea that chatbots are best suited as “get things done” services or virtual assistants. Though they certainly will fill that need, Livingston and Kik see a different possibility, one where bots, “will let you instantly interact with the world around you.” Livingston illustrates with the following example:
At a baseball game, he wants to order a beer. The stadium has developed an app. So instead of having to stand in line and risk missing any action, all he has to do is go to the app store, search for the app, put in his password, wait for the download, create an account, enter his credit card info, and then order his beer. As he recounts all of these steps, he realizes that this is no more convenient than leaving his seat and standing in line for his beer. What if, he wonders, he could simply chat with the stadium to order his beer in the messaging app that he already uses? This is where Kik is headed. In this scenario, Livingston sees a sticker on the chair ahead of him that says,“Want a beer? Chat with us!” On the sticker, there is a chat code. Livingston unlocks his phone, opens his chat app, scans the code, and is immediately interacting with the stadium to place his order.
But, of course, isn’t this what QR codes do? And, nobody uses QR codes, right? Livingston address this. He explains that QR codes uniformly delivered a poor user experience. Think of the last one you scanned, if you ever did. It is likely that you were directed to a slow-loading website where you had to dig to get to what you wanted. The chat code option keeps your interaction with the real world in the interface of the chat app that you are already familiar with.
“This all might seem too simple. But to us, that’s the beauty of bots. They can reduce friction to as close to zero as computing allows.”
Ted Livingston, Kik
Before your organization considers developing a mobile app, you should first consider the much morecost-effectivee approach suggested by Livingston and develop a chatbot for an already existing ecosystem. More on how to do just that, shortly.
I Love You, Xiaoice
You may recall the 2013 Academy Award-winning film, Her, written and directed by Spike Jonze. The film centers around a lovelorn introvert played by Joaquin Phoenix who becomes intimate with a talking operating system (OS) named Samantha. The film is rightly categorized as science-fiction, but many of the film’s themes are playing out in real life. As Joseph Weizenbaum found when users started interacting with ELIZA, our human desire for connection, intimacy, and empathy can, in fact, be satisfied by machines. In China, Microsoft has developed an empathetic chatbot and how users have responded has surprised them as much as ELIZA did Weizenbaum.
Xiaoice, or little Bing, more literally, “Microsoft Little Ice” is a natural language chatbot that Microsoft developed to engage users on Weibo, a Chinese micro-blogging platform akin to Twitter. Xiaoice replicates the personality of a 17-year old girl, and in contrast to typical chatbots, “she opposes users, offers independent opinions and conclusions, works to demonstrate caring, and is unpredictable, not always offering the same response to the same input.” Like many of her chatbot predecessors, Xiaoice passes the Turing Test; users can’t differentiate her responses from those of an actual human. Her spirited personality and balance of IQ (intelligence) and EQ (emotional intelligence) has won over more than 40 million registered users who engage in daily conversations with “her” that can last more than 26 turns. Let’s return to a Gartner’s key takeaway from earlier in the presentation: by 2020 “the average person will have more conversations with bots than with their spouse.” It becomes clear that humans are more than willing to engage with chatbots beyond the superficial to fill deep personal, emotional needs. Illustrating this is the finding that of Xiaoice’s 40 million users, 25%, that’s 10 million users, have messaged “her” to say, “I love you.” Clearly, the 2013 film, “Her,” is no longer science-fiction but the reality many are living today.
Order Flowers, Plan a Trip, & Get Tacos for Lunch
“We are trying to make [the customer service experience] more human. It’s easier to chat and text than sit on the phone.”
When Facebook announced that it was opening Facebook Messenger for developers to build chatbots, the example that Mark Zuckerberg used was 1-800-Flowers. He remarked, “It’s pretty ironic: To Order from 1-800-Flowers, you never have to call 1-800-Flowers again.” The company developed a linear chatbot (not allowing free-from conversation like Xiaoice) that allows users of Facebook Messenger to ask it to place an order or connect with a human customer service rep. After a couple of months of use, Chris McCann the president of 1-800-Flowers reported that 70% of the chatbot orders were from new customers and they skewed younger than their typical audience. He stated to his surprise that, “most customers, especially millennials, would rather interact with a robot than a human.”
Another early entry into the Facebook Messenger chatbot ecosystem was hotel booking site, Expedia. Proving that consumers may engage with chatbots for more considered purchases than flowers, the Expedia chatbot works more efficiently than either their website or mobile app channel. To engage, users simply compose a new message within Messenger to @Expedia. From there the bot will ask a few questions regarding destination city and date, then will reply with booking options right within the Messenger interface.
Facebook Messenger is not the only chatbot ecosystem in town. Several organizations have found success on the workforce messaging application, Slack. Taco Bell developed a chatbot, appropriately named Tacobot, that with a series of a few simple commands allows teams and individual workers to order tacos at their local taco bell.
As Purna Virji shared at her Inbound 2017 presentation, Taco Bell’s investment in the chatbot’s development was a mere $40,000. That nominal investment has yielded more than $10,000,000 in workplace taco orders to date!
Here at SilverTech our Internet of Things (IoT) Engineer, Shawn, has created several chatbots for Slack which have both brought our historic Ash St Schoolhouse headquarters to life and made our lives easier. One such innovation was his creation of Tempbot, a Slack enabled HVAC controller that we use to control the temperature in each room throughout the building.
Getting Started with Chatbots – It’s Easier Than You Think
Chatbots are essentially the “apps” of voice and messaging platforms, defining how people converse with your digital business services and data.
- Call center help desk
- ChatOps approvals
- Equipment diagnostic inventory management
- Chatbot scheduling agent
Further, chatbots could be used to turn your physical locations into beacons of engagements like the stadium in Livingston’s example. For example, Financial Institutions could develop a teller bot at the ATM that allows customers to ask questions that go beyond the ATM’s capabilities such as current loan and deposit rates. A hospital could develop a check-in bot that allows you to chat about wait times, pre-screening questions, and health tips. And, a university could develop campus bots that help students and prospects get around, find out dining hall wait times, and other useful campus info. The possibilities are truly limited only by your imagination.
And that’s where you should begin! Imagine the myriad ways in which your organization could use a chatbot. If you recall, it only cost Taco Bell $40,000 to generate $10,000,000 in orders. With the right chatbot, your organization could achieve comparable return-on-investment.
Practically, though, after you let your imagination run wild, there are steps that you can take today to prepare your organization for a chatbot. Even if you don’t have access to a team of developers and digital strategists (which is, of course, where an agency like SilverTech could help) a simple spreadsheet is a good place to start.
Connect with your sales team and/or customer service team, much like you did as you developed personas, and create a list of the most common questions that your audience asks your organization. Then, in that same spreadsheet, use a drop-down to list all of the possible answers. This is the logic that your development team will use to build the chatbot.
If you already have an FAQ section on your website, you may have already undertaken this first step. An FAQ chatbot is a great way to enter this channel. In fact, Microsoft offers a platform, QnA Maker, that enables organizations to “build, train and publish a simple question and answer bot based on FAQ URLs, structured documents or editorial content in minutes.”
As we enter 2018 and head toward a reality in which this approach is the norm, it is returning to Purna Virji’s advice, “Caution is necessary, avoidance is not.”