Beware the Corporate Catfish
Three things brands should look for in a conversational AI
Welcome to the broken-hearts club.
Does anyone remember Manti Te’o? Ten years ago he was the best thing happening in college football. A linebacker at Notre Dame known for his dominant tackling and an ability to score on defense, Manti was an NCAA darling. Off the field, he was an Eagle Scout and named a National Scholar-Athlete. On the field he won just about every award a defender could win while also placing second on the Heisman Trophy list; only one other defense-only player in history had made it that high on the list. Manti still might be the best college football defensive player ever but does anyone remember his 21 tackles in a single game against Stanford? No. Manti Te’o is remembered for getting famously, and tragically, catfished.
At the climax of his historic college football career, a devastated Manti Te’o announced that his mother had passed away and that his girlfriend, Lennay Kekua, was dying of cancer. The story of his girlfriend was gut-wrenching. She had been in a car accident and, during her hospital stay, had discovered that she was in the advanced stages of leukemia. Devoted and selfless, Lennay told Manti that instead of coming to the hospital he should focus on football and play in her honor. The story read like a movie script and, being the superstar he was, Manti played like a maniac for his dying girlfriend. The most tragic part of Lennay’s story was not her cancer or her early death, it was that she never existed at all. Lennay was a character invented by a man named Ronaiah Tuiasosopo. Ronaiah had fallen in love with Manti and had created Lennay as a way to get close to him. The season-long love story of the year was a huge embarrassment and launched both Manti and the term “catfish” into the national spotlight.
Making friends is easy.
To get catfished means to get tricked into believing a fake online persona is a real person. Manti Ta’o’s story is terrifying because he was a smart guy; it’s terrifying because catfishing is so easy to do. In a public statement, Manti describes just how easy it was. “This is incredibly embarrassing to talk about but over an extended period of time, I developed an emotional relationship with a woman I met online. We maintained what I thought to be an authentic relationship by communicating frequently online and on the phone, and I grew to care deeply about her.” The digital love affair between Manti and Lennay was not love-at-first-DM, it required time and channels to develop. First, they communicated “frequently”. Lennay asserted her presence multiple times a day, reinforcing her availability. Second, they communicated “over an extended period of time.” The longer the communication was sustained the deeper Lennay established a familiar pattern with Manti. Third, Lennay showed up in more than one place, communicating “online and on the phone.” From Twitter to text message to email, each new channel she used was validation that Lennay was real and interested in Manti. In a classic episode of Star Trek: The Next Generation, Data gives us his dispassionate description of how attachment develops, “Friendship is sometimes less an emotional response and more a sense of familiarity… As I experience certain sensory input patterns my mental pathways become accustomed to them. The input is eventually anticipated and even missed when absent.”
We are all swimming in a sea of sensory input from our news feeds, inboxes, notifications, and messages. We rely on LinkedIn to network with other professionals, we rely on Instagram to find people that are doing interesting things, and we rely on dating apps to meet new love interests. Everywhere we turn we are exchanging information with people that we’ll never meet in person. Where patterns of communication emerge we begin to develop expectations, trust, and even adoration. We gravitate to the people that show up often, that stick around, and that connect with us in multiple ways — time and channels. I believe that most of us do our best to represent ourselves accurately and to connect with people that are honest about who they are. Emerging tools are giving brands the opportunity to build on-line relationships with massive numbers of people and the broader their reach the greater the imperative to show up in an authentic way.
Conversational AI is coming and it could be great.
Our virtual ecosystem provides a natural habitat for virtual agents. These agents settle in among the humans, seamlessly performing tasks we would otherwise find tedious or difficult. Many companies, especially those that are large or growing rapidly, are already finding uses for virtual agents to engage customers and employees. Their agents trigger drip campaigns, place banner ads, and manage social media spends. While this work can help marketing teams operate more efficiently it amounts to little more than talking at people. Conversational AI is around the corner that will revolutionize the way we engage customers and employees by talking with them, just imagine your brand slipping into the DMs of thousands of people at the same time. The current generation of chatbots can send us messages and may even accept responses but they are limited to using natural language processing to understand what people say. To take on direct engagement conversational AI will have to understand what people mean. An AI that can communicate with its target audience frequently, over an extended period of time, and across multiple channels will undoubtedly build rapport and even trust with its target audience. Sounds great right? Let’s not forget: no one wants to get catfished. So here lies our challenge, how do we deploy conversational AI that’s engaging but not deceiving?
Excellent communication starts with excellent listening.
A virtual agent’s ability to understand like a human is much more important than its ability to talk like a human. If I were to ask you to tell me about the last conversation you had with your boss you would probably tell me when it was, where it took place, how you were feeling, what your boss was wearing, the tone of their voice, who had scheduled the meeting and a dozen other details besides what was actually said. These details about context provide clues about intent. The more effectively a chatbot comprehends intent the more it will begin to resemble conversational AI. For example, the typical chatbot sitting on a corporate website is more like an interactive FAQ, it doesn’t let me skim through topics and only gives me the right answer if I use the right words. If, however, that bot were to observe how I have navigated to the site, what I have clicked on, and where my mouse has hovered it could use those context clues to infer intent and refine it’s responses accordingly. The ability to utilize context will be even more important as your virtual agents reach outside of the curated environment of your corporate sites and into the wilderness of your audience’s personal channels. Introductions on LinkedIn are very different than on Twitter or Tinder and what is considered an appropriate response in one context may be wildly inappropriate in another. The survival and success of your conversational AI will hinge on its comprehension of context and intent.
Understanding context and intent will not be easy to achieve, hell, humans struggle at this. How many times has an email chain spun out of control because someone read the wrong tone into the message? Comprehension is the biggest technical hurdle preventing conversational AI from reaching its potential as an engagement tool. Whether you are developing a virtual agent inhouse or are in the market to buy one, you should focus the lion’s share of investment on comprehension. If a vendor doesn’t have a clear roadmap with significant investment in the future comprehension capabilities of their bot then don’t buy that bot. Excellent listening is the core product, the rest is features.
Invest early in conversational design.
To compensate for a bot’s limited comprehension and capabilities, product teams often turn to design. Excellent UI design can generate interest in a product despite limited features so it’s only natural that these teams would focus on the look of the bot, giving it a clever name and maybe a human face. This skeuomorphic tactic gives users the subtle queue that they can try speaking to the bot like a person. However, the anthropomorphization of your bot will only frustrate your users if it can’t actually understand like a human. Part of the problem here is that the user interface isn’t the buttons or windows, it’s the conversation itself. A conversational AI will be talking to your audience across many channels in which you won’t control palettes or fonts or button placement. Once you’ve established a roadmap for comprehension your next priority is excellent conversational design. The team at Google has published an excellent example of a conversational design guide for their Google Assistant. If you are building or buying I strongly recommend using a guide like this one to assess whether or not you are looking at a product that will promote early engagement as you work to add features. Your goal isn’t for people to remember your chatbot, it’s to have them remember your brand. Conversational design can help focus your audience’s attention where it belongs.
The best name for your bot is your brand.
Speaking of Google Assistant, Google also provides an excellent example of how to name a conversational AI. Instead of wasting calories with clever names and handsome avatars, Google has made their bot their brand. Yes, there are some examples where the name of a bot has become synonymous with a brand (Alexa and Siri) but this has taken years and significant investment to sell these assistants as brands within themselves. If you’re looking to use conversational AI to promote your brand or products or jobs then the bot is not the goal, the engagement is. The most straightforward solution here is to simply give the bot the name of your brand. There’s no reason to force your audience to remember a name they don’t need. Beyond the benefits of a clear brand message, this approach also neatly sidesteps the real risk of catfishing your audience. As conversational AI gets more fluid and humanlike you’ll have to be more intentional about letting people know they are talking to an entity and not a person; none of your customers are going to high-five you for acing the Turing Test.
The future is brands that network like humans.
I love meeting new people. Over the years some of my most meaningful friendships have been built and maintained online. The further I get into my career the more I rely on my digital presence to foster new friendships and professional contacts. Wanting to put my best foot forward I carefully curate my online persona and make sure that I have pictures that show me as I am and as someone worth knowing. As much time as I have spent thinking about how I appear, my personal brand is really built on how I engage. Sending kudos to colleagues, adding meaningful comments, and contributing articles are activities that build trust with my network. As I have engaged frequently, over a long period of time, and across multiple channels, that network has grown. The last thing I want to do is catfish anyone and undermine my network. As conversational AI continues to develop your brand will have an opportunity to network with humans like a human. This capability comes with a warning and a responsibility to not make the mistake of pretending to be human. Don’t trick people, talk to them. Focus on excellent listening and engaging conversational design and you will see your brand’s network grow and be trusted.
