ChatBots: A Love Story
The last twelve months have been awash with talk of artificial intelligence and automation disrupting our lives. As often happens with new, hyped technology, the beginnings of a backlash are emerging, with many industry analysts and general consumers scratching their heads and asking “well, where’s my genius robot?”¹
There are two typical responses to this line of thought.
The first is that AI will not just ‘click’ into existence. Rather, it will slip into existence. Much the same way a human child develops incrementally, before one day — almost surprisingly — emerging as a fully functioning adult, AI too will slowly emerge — bit by literal bit. There won’t be a day when the switch is flicked and our personal robot-assistants arrive. They’ll just be here one day before we realise it.
The second response is that AI is materialising in ways we simply didn’t imagine, from ambient-computing gadgets like the Amazon Echo, to the advanced algorithms providing us with our Google search results. Though work is being done — primarily in Japan — towards the classic humanoid-bots often represented in pop-culture, there’s another kind of AI that is steadily materialising before our eyes across multiple industries.²
In fact, if you’ve done any online banking today, ordered a pizza online, or chatted to a work colleague over social media, you might have engaged with this other kind of AI and not even known it.
The ChatBot Revolution
“ChatBots” are digital assistants that respond to natural language queries from users (typed or spoken) in a responsively natural way. Think Siri, Google Assistant, and Amazon Alexa, amongst others. Most ChatBots take the form of a modal-window on a webpage. But they can vary. A Facebook company page that responds to messages sent over Facebook Messenger in a human-like way is the most common form of ChatBot.
The rapid emergence of these types of ChatBots is being driven by the low cost of entry. Most of these ChatBots harness open-source technology. The primary investment required is not in the tech, but in the business logic — and there’s a lot of business logic pointing to the benefits of ChatBots.³
To understand their rise, consumer-appeal, and return-on-investment, it’s necessary to venture briefly into the field that presides over all market innovation; human psychology.
The Psychology of conversational UI
Historically, people have always mentally attributed human-like characteristics to all sorts of entities that aren’t human. Whether object or animal, we’re not afraid to cognitively stretch our perceptions to see, believe, and interpret human traits if there’s even the slightest indication of human-ness.⁴
Ever get tempted to yell at your computer, explain to your dog why their dinner is late, or feel that an expensive car has an arrogant ‘face’? This is called ‘anthropomorphism.’ The computer, in all its forms, with all its promises of ever-increasing processing power, has always been a desired target of this anthropomorphic attribution.
Most ChatBots are relatively simple under the covers, often simply looking for key words mapped to canned responses.⁵ Counter-intuitively, predictability is often a positive thing for anthropomorphism. Dealing with true, complex social interaction requires cognitive effort; decoding non-verbal cues, listening to new information, adjusting your communication and being timely in your response. A ChatBot doesn’t require any of this cognitive ‘cost’ — yet provides nearly all the social and informational benefit as true interaction does. Because, as humans, our brains are prone to laziness, we’re often perfectly willing to believe that the bot we’re interacting with is human — just like us.⁶
In essence, ChatBots are a new but fundamentally traditional medium of interaction with a computing system. Users converse with it, rather than program it. Designers call this a Conversational User-Interface (CUI) — a term that’s getting as popular for developers behind the scenes as ‘ChatBot’ is for end-users.
Product psychologist Nir Eyal looks at factors influencing the likelihood that products will result in ‘habit-forming.’ Eyal theorises that if you design a system responsive and interactive to pre-existing skills, it is far more likely to get used — and used repeatedly — than a system that requires the development of new modes of interaction⁷. For example, studies demonstrate that most knowledge workers spend around thirty percent of their day searching for information. Products featuring conversation user-interfaces — like ChatBots — mean these workers can obtain information through their natural language capacity, without the obstacle of developing a new skill.
That’s good news and bad news. On the one hand, the most efficient way of interacting with a tool is not through CUI. Imagine having to say “open Google Chrome, go to salesforce dot com, login in with my email, go to reports.” While even these tasks will eventually be conversationally ‘clumped’ into more condensed forms, with AI extrapolating the components of a complex task, it’s worth remembering that the big players — Google, Amazon, Apple — aren’t there yet. Neither are the small-to-medium businesses.
Eyal and others warn of an over-focus on ‘ultimate-efficiency’ vs ‘right-now-efficiency’. Does the superior product always win? No. The market-adjusted one usually does. Consider the following graphs:
In a perfect world, the technology of the green line is preferable. Given enough time, it will lead to higher productivity. But in our age of ‘app-for-that’ thinking, subscription product access, and hyper-competition, a crucial metric that must be considered is the concept of the user’s ‘willing time investment.’ Put another way, if the user is unable to achieve a level of productivity with a tool within a certain time, they will abandon it altogether.
On the ground, companies often fall into the trap of product-myopia — becoming so indoctrinated with the perfect solution for a market ‘pain-point’ that they forget what it’s like for a real-life customer to use the product.⁸ Driving this myopia is a misconception about the importance of the company’s product in their users’ lives. The product usually represents close to 100% of a company’s daily attention. For the user, the product is typically closer to 1% — even with high-involvement products.
So perhaps the way you’ve structured your product-interface is the most efficient and effective. In a perfect world, all customers have conquered the learning curve to become perfect users. But real-world designers cannot underestimate the reality of the learning curve. They must consider the very real possibility that users of the product never will, nor ever even intend to, conquer the green line. In the real world, the time investment versus return simply doesn’t match up.
Let’s look at that graph again with this idea overlaid:
Long-term, there doesn’t always have to be a trade-off. The trick is not overestimating or underestimating the time or attention-span of your users — but rather, catering to it. Conquering the inside of the red line to ensure a healthy user base, while working to deliver whatever magic lies at the top end of the green line, within the time constraints of the blue line — is the key to success.
In most competitive industries, one thing is certain. If you don’t do this, a competitor eventually will. In fact, in many industries, competitors already are.
Rise of the ChatBots
In the late 90s, large telcos with big support departments began answering their customer’s phone calls with pre-programmed voices aimed at sorting and filtering the calls. Commonly, users were also directed to seek help through packaged manuals and the internet, rather than continue the call. Were these early Chabots? Yep. Did they win hearts? Nope. These efforts — primarily driven at lowering cost, rather than increasing value to customers — created customer interactions that were simplistic, ‘lesser-human’ and essentially, not fit for purpose.⁹
Domino’s Pizza has spent the better part of this decade becoming the food-industry darling of the tech world’s eye. Some experts have even proclaimed that the company is now a technology one that happens to serve food, rather than a food company using technology.¹⁰ From their early entry into the app-marketplace, to their splashy drone-delivery tests, Domino’s has cultivated a competitive advantage in technology and seen its market share rise commensurately. Enter the Dominos ChatBot.
The bot, anthropomorphically named “Dom”, is described by the company as “an artificially intelligent customer whiz designed to help superfans get their number 1 fix of cheesy food heaven.”
Remember that graph? Tellingly, “Dom” was developed within Facebook’s messenger platform, rather than on the organisation’s own app, even though this meant sacrificing technological control and prowess. Facebook — with over one billion active users — means ‘Dom’ enters a massive pre-existing market, rather than the comparatively minuscule market of Domino’s customer base. Better yet, people already know how to use Facebook Messenger, so there’s no learning curve.
In the world of suits-and-tie, LinkedIn has entered the game with a ChatBot of their own, available now on the desktop version of their website. The LinkedIn Bot actively monitors your messaging with acquaintances, suggesting conversation ice-breakers, ways to frame discussions, and strategies of acquiring meetings with important people.¹¹ It’s a great solution for potential employees have who have the right skills, but not the small-talk to get their foot in the door. Other brands developing and enjoying success with Chatbots include Starbucks, British Airways, and eBay.¹²
Other industries are starting to take note.
ChatBots for Training Providers
The training and education industry offers one of the biggest opportunities for ChatBots to shine bright. With data-points (in the form of courses, classes, and assignments) often numbering in the hundreds of thousands, finding answers to common questions like “what design-related classes can I take on Tuesdays” are perfectly suited to a ChatBot response.
Tech startup AdmitHub, which since 2016 has provided a text-message interface for potential students seeking course and admission information with universities, has already inked deals with Georgia State University, BSU, The Cooper Union, West Texas A&M, and others.¹³ Stanford’s in-house built ChatBot, called ‘Stanford Jane’, has been actively used and updated for years. Further upmarket, Ivy.ai provides a similar solution, albeit focusing on the deep-learning and technical AI aspects of ChatBots. Clients include Penn State, Auburn University, and the University of Virginia.¹⁴
Training companies looking to gain a competitive edge with technology would do well to place ChatBots at the top of their priority list, think ‘extension’ and not ‘replacement’, to focus on usability over capability, and most of all — to get creative.
References:
1. http://www.wired.co.uk/article/sensationalism-ai-hype-innovation
3. https://venturebeat.com/2016/10/18/3-ways-chatbots-deliver-roi-for-customer-service/
4. https://www.psychologytoday.com/blog/behind-online-behavior/201607/the-psychology-chatbots
5. https://www.forbes.com/sites/parmyolson/2016/02/23/chat-bots-facebook-telegram-wechat/#3110669d2068
6. https://medium.com/@botanalytics/understanding-psychology-behind-chatbots-51d9614a5042
8. https://hbr.org/2004/07/marketing-myopia
9. https://www.uipath.com/blog/rpa-and-telecommunications
10. http://www.afr.com/business/retail/fmcg/forget-pizza-dominos-is-a-tech-company-20150810-giw8dj
11. https://techcrunch.com/2017/01/19/linkedin-desktop-update/
12. https://chatbotsmagazine.com/3-reasons-2017-is-the-year-of-the-chatbot-6fa0c783a444
13. http://blog.admithub.com/case-study-how-admithub-is-freezing-summer-melt-at-georgia-state-university
