3 trends driving the chatbot revolution
By David Temkin for Venture Beat
Facebook’s big announcement at its annual F8 Developer conference, was to make Messenger, its 900-million-user messaging app into a full-fledged platform that allows businesses to communicate with users via chatbots. On the surface, that sounds like a standard, if very ambitious, product announcement, taking an already successful product to the next level. But it’s a lot more than that, and it’s worth looking into the three intersecting trends rolled into this one announcement to understand why it’s a big deal, and why the the chatter around chatbots is everywhere.
The three trends:
- Messaging-as-OS: Messaging can be a platform
- The app problem: People are reluctant to install apps
- The “conversational interface”: A new model for interacting with online services
What we quaintly call “phones” aren’t used much for phone calls these days. In fact, measured by actual usage, they’re messaging devices; for most users, messaging (texting and all of its more advanced variants) is the core function of the phone, the place they’re focused on most of the time. So: If users are already spending most of their time in messaging apps, and have gotten very comfortable with how messaging works, then why not bring more and more functions right into messaging apps, rather than asking users to pogo-stick from app to app, from site to site? Go where the users already are.
Of the three major trends in Facebook’s announcement, this one is standing on the most solid ground, largely because it is the least driven by internal tech industry obsessions. Not only is messaging demonstrably the center of today’s smartphone usage, but if you look to Asia, you’ll see an alternate set of messaging apps dominating the landscape, all of which do more than their Western competitors, and some of which have a legitimate claim on being a “Messaging OS” right now. Most notable of all is WeChat, an app that’s widely used in China, with 570 millions users and over a million businesses represented in the app. You can use WeChat to chat, of course, but you can also order a meal, book a doctor’s appointment, donate to a charity, and more. Not only can you do these things, but they’re everyday events for users of WeChat. WeChat connects people to people, but also people to businesses, and this has become totally mainstream for users of WeChat. So much so, in fact, that new companies launching online services will sometimes launch first in WeChat, and only afterward produce their own app.
That’s why you’ll hear people talk about “the WeChat of the West” — what would it take for people outside of Asia to to use their messaging apps this way? It’s a tricky question. But despite the fact that WeChat allows people to do way more than chat with other people, WeChat doesn’t force all of these interactions into chats with bots. When you interact with a business in WeChat, much of interaction is done using embedded mobile websites inside the WeChat environment. So you wouldn’t chat with a bot using keywords to book an appointment. Instead you tap your way through a process that closely resembles what you’d do in using the web or a dedicated mobile app; it’s just that you don’t need to leave WeChat to do so, and you don’t need to download a new app or register for a new service to get things done. It happens on-demand inside of WeChat.
Solving the app problem
Businesses want customers, and customers are mobile. The best way for a customer to interact with a business on a smartphone is generally with an app. But good apps are hard to build, and even harder to get into users’ hands. The average smartphone owner installs zero apps per month, and only uses 5-10 apps regularly. And to add to that, there are around three million apps available across the two major platforms — it’s overwhelming. People don’t much like loading up their phone with apps, and are tired of trying out the latest-greatest app. So what’s a business to do if it wants to interact with its customers in a mobile world? Some businesses have the scale to develop a custom app and drive it into user’s hands. But many — most — don’t.
One of the big attractions of Messenger’s chatbot platform is the idea that customers would be able to interact with a business without downloading an app — or even without the nearly-as-awkward process of going to a website, logging in and, if needed, manually entering payment information. There’s talk of “bot stores” and the idea that there could soon be millions of different bots, just like there are now millions of apps — but they may well prove easier to build, and easier to get in front of users.
However, it’s an open question — a very open question — whether bots are an answer to the app problem from the perspective of a user. Can a bot really act as a substitute for an app? Which leads to…
The emergence of the conversational interface
This is the the biggest and riskiest idea that Facebook is promoting — the idea that people will want to interact with businesses as if they are people, in conversational format. Two ways to look at this:
The command line makes a comeback. You might just remember what computers were like before the advent of the GUI. Users would interact with the computer via a command line, which required memorizing a series of often cryptic commands in order to get anything done. To this day, developers tend to interact with computers in this manner, because it can be very powerful and very flexible for those who invest the effort to master the technique. But for normals, it’s an absolute show-stopper, and without the GUI, computers and the Internet would never have taken off. Which brings us to bots. In their simplest variant, a bot offers a classic command line interface. You need to know what words to use in order to get the right response; and instead of making selections from menus and doing all of the usual touch-swipe interaction on a smartphone, you’d have to enter textual commands — specific sets of words, delivered in a specific order — that are known to the bot.
One innovation that Facebook showed at F8 is called structured templates. So when you’re conversing with a bot in Messenger, it might respond to you with something more than plain text. A bot might send you an image with a few tappable options. That might just beat having to interact along the lines of a phone tree (“press one to buy shirt one, two for shirt two…”). Might. It feels like a stretch, or a partial solution that compensates for the lack of a standard graphical interface.
The real excitement around bots, though, isn’t a drive to restore computing to the glorious command line era. The promise is instead to take a bold leap into the future, in which interacting with a digital service will be more like talking to a person — artificial intelligence in the HAL 9000 sense of the term. What if you didn’t need to use a special vocabulary, and were able to interact with a computer system — a retailer, an airline, a doctor’s scheduling system — using natural language, the way you’d talk if you were talking to a human being? This is the (potential) aspect of bots that’s captured the imagination of the industry. Now, the promise of intelligent agents doing your bidding via normal conversation is an old dream, going back to the earliest days of sci-fi. Periodically the industry focuses on the subject of AI, and up to this point, each attempt to create conversational agents that can pass as human beings has fallen short. Even with mainstream services like Siri, there’s a giant chasm between the original magical promise and the awkward and not-entirely-useable reality. To put it a different way: How useful would your iPhone be if it didn’t have a screen, and only had Siri?
When you put it all together, a few things become clear. First, Facebook has taken a big leap into a domain that isn’t proven. Conversational interfaces are not something that have yet succeeded in the mass market, and are not something that users are clamoring for. Yes, other messaging platforms (notably WeChat) have gotten very far by integrating many services into messaging, but not by building out bots. Instead, WeChat and similar platforms typically use the proven format of the GUI, in the familiar form of mobile websites embedded in the app.
Facebook and the other industry players pushing conversational interfaces (Microsoft, Amazon) are driving computing into a new era that is at once retro (a command line, or even a phone tree) and simultaneously futuristic (artificial intelligence). But, this is a tech-driven vision, not one driven by a concrete user need or demand. Users aren’t clamoring to book appointments or buy shoes conversationally. And the same is true of the app distribution issue. Users may not like the idea of not having to download apps, but they aren’t saying that they don’t like apps. They love apps. They spend all day in apps of one kind or another (including messaging apps!).
Every so often the tech industry gets obsessed with an idea that isn’t driven by concrete user needs or problems. The conversational interface might just be one of them. On the other hand… you never know what kind of innovation will be unleashed by thousands of developers building on a platform that reaches nearly a billion people. We may just be surprised.
David Temkin is the CEO and co-founder of Cola, a company that offers a messaging OS to make texting more actionable and efficient
First appeared in VB