The app boom is over. There are now more than 4.2 million apps available for Android and iOS, but three-quarters of American smartphone users download a grand total of zero new apps per month. They might be mostly free and easy to access, but apps are struggling to make it on to our phones and tablets. According to comScore, we spend the majority of our screen time using just three apps, with the average American spending almost half their time in just one.
The most popular type of app? Messaging. With the eyeballs of the world glued to WhatsApp, Facebook Messenger, WeChat and Skype, developers have started turning once-simple chat apps into complex ecosystems. And at the centre of this change is a horde of subservient bots.
It's Friday, November 2, 2018. You've just walked into your kitchen after a long week. "Play Etta James," you say. "At Last" starts playing throughout the house. Your phone vibrates in your pocket; a notification on the screen reads "Return flight to Toronto now £220 per person. Book?" You type "Yes" and confirm your identity using the phone's thumbprint reader. You open the kitchen cupboard and scan for ingredients to cook before returning to your phone and entering two emoji: pizza, wine. Twenty minutes later, the doorbell rings.
Think of the last interaction you had with a software application. You likely used a mouse, keyboard or touchscreen to navigate a series of options to complete a task. In doing so you were forced to follow the rules of software laid down more than 30 years ago; it is the software that dictates the rules of engagement, not you. Now, thanks to bots, those rules are changing.
On chat platforms in China, government services in Singapore and speech-based personal assistants in the UK, bots are taking over. The next user interface won't be based on skeuomorphic design or muddled menus; it will be based on simple conversation. From Slack to WeChat, Kik to Facebook Messenger, and Telegram to Amazon Alexa, bots are becoming the main interface between humans and machines."There's a huge opportunity here," says former Evernote CEO Phil Libin. "Within a few years bots will be in the fabric of everything."
Libin, who is now managing director at the Cambridge, Massachusetts-based venture capital fund General Catalyst, is investing in and working with startups that he believes will take bots mainstream. The bots he's interested in are not intelligent, human-like assistants; they are far simpler. He refers to them as a "conversational interface", a means of interacting with software through speech, text, emojis, images, video or other means. "You don't have to learn how to use it, you just use it," he says. "You have a conversation and it lets you do whatever you want without having to navigate through a stack of options."
For writer and publisher Tim O'Reilly, the emergence of bots harks back to the earliest days of computing. "As we move into conversational interfaces we really are moving back into the world of the command line," he explains. The limitations of the command line, with its reliance on explicit prompts, soon saw it usurped. Despite its limitations it remains the simplest way of interacting with a computer. After all, what's more natural than having a conversation?
"Look at the long history of user interface as a convergence of what machines are able to understand and how much work humans have to do to help them do so," says O'Reilly. "In the old days it had to be very, very explicit but now it's becoming less and less explicit because there's more fuzziness in what the machines are able to handle."
What exactly a bot is depends on whom you ask. For some developers and investors, a bot is a sophisticated, artificial-intelligence-infused creation capable of understanding pretty much any interaction. For others, it is an intentionally dumb interface, capable of understanding only a limited number of predetermined commands. Whereas the former seeks to ace the Turing test by engaging a user in a conversation, the latter might simply ask, "What pizza would you like? Press one for Margherita, two for Diavolo, three for Hawaiian and four for Quattro Formaggi".
O'Reilly describes bots as having "a little sprinkling of AI" rather than relying on a layer of artificial intelligence. "If you look at a bunch of bot toolkits they're really just much smarter versions of branching," he argues. Libin agrees: "There are some misconceptions in the popular imagination about what this is going to be like," he says.
"Probably the main one is that bots should talk like people. If I engage with a bot to do something, I don't expect it to behave like a human, I don't expect it to talk to me as a person would. I expect it to talk to me in a way that's much better than what a person would. I don't want to mimic a human experience - I want to have a much better experience."
One of Libin's investments is Growbot, a messaging bot that listens for and encourages praise on Slack. "The bot is a participant in the conversation that adds structure and functionality," he explains. When Growbot spots praise it reacts, keeps a tally of who's saying what and compiles a report for managers. The company has raised $1.7 million (£1.3m) in two rounds of seed funding and is used by more than 2,000 companies, from Starbucks to London-based advertising agency Spongecell.
Slack, with its focus on teams at work, has become an early pacesetter in an industry still searching for its killer product. In July, it announced a $2 million investment in 14 startups working on bots for its platform. The money is part of a bigger $80 million investment vehicle announced in December 2015 featuring Accel, Andreessen Horowitz and Index Ventures. Since launching in August 2013, Slack's growth has been rapid. It has three million daily active users and 930,000 paying subscribers.
"We have some bots that totally reside within Slack," says April Underwood, vice president of product at Slack. These bots, she explains, are helping people to complete irritating tasks that aren't core to their job: file expense reports, get budget approval or order new office supplies. "Slack allows bots to join the conversation and solve those tasks in a quick way from the application teams that are already in."
Where Slack has already gone, others will follow. "The kind of behaviours that you see in Slack are going to be fundamental to all the Microsoft platforms," argues O'Reilly. "It's how you're going to invoke actions on a computer - whether that's a typed or a spoken conversation probably doesn't matter." A bet against bots would, according to O'Reilly, be "totally stupid".
For bots on Slack, simplicity is key. "A lot of these bots right now don't have to be super sophisticated. They can start from very simple commands," says Underwood. This means you don't have a conversation with a bot - you just tell it what to do. "Think of them not as people but almost like service animals," says Libin. "A sheepdog is super-humanly good at its job: it's going to herd sheep much better than any person could. But it isn't clever. It's something that is fantastically good, much better than any human could possibly be."
Take Envoy, a Slack bot that sends direct messages to employees whenever someone arrives at the office to see them. It solves a simple, non-core job task in a simple way, without staff needing to install and learn a new system. Founded in San Francisco in 2013, Envoy has raised $20.31 million in three rounds from investors including Reddit co-founder and executive chairman Alexis Ohanian, and Jeremy Stoppelman, co-founder and CEO of Yelp. "If we can do that a thousand times over for all the things that need to happen inside the typical workplace every day - then I do think the experience of being a worker is going to get better and better," Underwood says.
Outside the workplace, bots have even greater potential to change how we engage with technology. "A lot of people I interact with believe they now type more words than they speak," says Sarah Guo, an investor working on the enterprise team at Greylock Partners. "That's a fundamental shift, for most of our communications to be digitally captured. And yet most of our software today doesn't take advantage of the rich data we are creating in our constant communications, instead requiring us to do structured, unnatural data entry."
You've been unwittingly interacting with a bot since 1998 - it's called Google search. So why has software been so stubbornly skeuomorphic? Why should ordering a pizza involve downloading an app, signing up for an account and then finding the menu option for extra chillis? The great promise of bots is that they will break down the stubborn barrier between human and machine and make scores of apps redundant. According to O'Reilly, the switch to conversational interfaces will be rapid: "Will I be pawing at the screen of a 2019 Tesla? No, I won't. I'll be talking to it."
In the space of just three months, Facebook, Google and Microsoft collectively fired the starting pistol in the next big platform race. In March, Microsoft launched and open-sourced its developer tools for making bots; in April, Facebook announced its own bot developer platform based around its Messenger app; and in May, Google showed off Assistant, a new AI personal assistant. But the west is playing catch-up. Messaging platforms in China, unconstrained by an established app economy, are already showing the way.
Weixin, the Chinese name for Tencent's WeChat app, enables 762 million people a month to book taxis, check in for flights, buy concert tickets, donate to charity and play games without opening another app. Bots are also bridging the gap between the physical and digital worlds: Chinese restaurants are combining Weixin bots and QR codes to let customers browse menus, order food and pay - all by chatting to a bot.
"It becomes the new interface to the offline world," says Ted Livingston, CEO and founder of messaging app Kik. "There are now more bots put on Weixin every day than there are websites put on the internet in China," he explains. "In other words, Weixin is the internet in China." Kik, which has partnered with and received a $50m investment from Weixin owner Tencent, is now one of several western companies taking lessons from China's burgeoning bot economy.
In April, Kik launched its own bot store, which has already attracted more than 20,000 bots serving CNN, Victoria's Secret and H&M. To date, more than 1.8 billion messages have been exchanged with bots on the chat app, with users who spend 32 per cent more time chatting than non-bot users. Bots provide style advice, help with bra fittings and give weather updates. "The potential is limitless," Livingston says. "Chat apps are the new browsers and bots are the new websites." Despite the hyperbole, he admits it's early days. But the bot industry's development will be rapid.
"During the next six to 12 months I think you're going to see the release of the next wave of bot ideas that are actually well thought through and well designed and intended to be taken seriously," says Libin. The current limitations of bots are clear, agrees Livingston, but better developer tools will help build richer experiences."Today it's largely text-driven," he says. "I think about it like the browser was 20 years ago; there was just a bunch of multicoloured text and then people added pictures, videos and elements that can move. Fast-forward 15 years and you can build any app you want in a browser. Chat apps will go through a similar progression."
To make the shift to transformative technology, bots need a substantial back-end. "There are going to be big cloud platforms that will deliver the fundamental intelligence that makes a lot of bots possible," says O'Reilly. Step forward Apple's Siri, Google's Now and Microsoft's Cortana, all the result of significant investment in artificial intelligence and natural language processing. The huge quantities of data already collected by these platforms will underpin conversational user interfaces as they spread throughout technology.
"If I were Google Cloud Platform or Cortana, I would be out there going, 'OK, how are we going to speech-enable every device that's out there in the world?'" O'Reilly says. "There's going to be a lot of platform-level functionality from the big players, which is going to get better and better. And it will include more and more AI. You already have Azure Machine Learning and Google Cloud Machine Learning and things like that. A lot of things that are complex today are going to be solved by platforms."
"Bots are the new apps," Microsoft CEO Satya Nadella declared in March 2016. Of all the major technology companies, the Seattle-based firm has made the biggest statement of intent on bots. Its open-source Bot Builder is a gamble, but one that could enable to it become the leading provider of the "fundamental intelligence" posited by O'Reilly.
So what's in it for Microsoft? Whenever a user interacts with a bot built on its platform, even if it is deployed on Slack, Facebook Messenger or elsewhere, Microsoft's AI gets smarter. "The more traffic we see on our system, the more intelligent it becomes," says Derrick Connell, corporate vice president of Microsoft's Bing division.
With its new bot obsession, Microsoft is also looking to China, where it has scored an unlikely success of its own. Since it launched in May 2014, more than 40 million people have held more than 20 billion conversations with Xiaoice, its artificial-intelligence-powered chatbot. "We started with a theory: can we maintain a conversation with another human?" says Connell, who also heads the engineering team that powers the bot. A key measure of the bot's success was how long it could keep the conversation going. "With our first version we were at 12 conversations per session. Three years later we're now up to 23 on average." The bot lives on Weixin and can now understand text, images, video and voice.
In December 2015, Xiaoice's familiar female voice started presenting the morning weather on the popular news channel Dragon TV. She's since moved on to reading the news and has fronted the channel's 2016 Olympics coverage.
Microsoft's other big bot experiment, Tay, was less successful. The bot, based on Xiaoice, was designed to mimic the language patterns of a 19-year-old American girl and learn from interacting with humans on Twitter. Released on March 23, 2016, within a day Tay - having learned from the worst social-media has to offer - was spouting racist and sexually violent messages. Two days later, after more than 96,000 offensive tweets were deleted from Tay, the experiment was taken offline.
Undeterred, in July Microsoft announced a partnership with the Singapore government to develop a bot to handle public services. "If you're a citizen of Singapore you can interact with a bot that works on behalf of the Singapore government. It can answer questions, you can register complaints, you can interact with a bot representative of the government," explains Connell.
In the west, it's Amazon that has had the surprise success. Amazon Echo, launched in November 2014, uses a conversational interface called Alexa. Want to listen to Daft Punk? Say, "Alexa, play Daft Punk". Want to set a timer to boil an egg while the music is playing? Say, "Alexa, set a timer for four minutes". Want to order an Uber? Say, "Alexa, ask Uber to request a ride". Amazon's bot is always listening and can understand commands. In June 2015, Amazon opened up Echo to developers - within a year, 1,400 new "skills" (Amazon's jargon for "app") had added support.
Unlike Microsoft's Xiaoice, which tries to imitate human interaction, Echo follows the same "dumb" bot principles laid down by Weixin. You're not going to have a conversation with it, you just tell it what to do. "Speech is more game-changing than people realise," argues O'Reilly. "It's been around a long time with Siri and Google Now and all kinds of applications powered by Nuance, but, to my mind, Alexa and Amazon Echo are game-changing in the same way that the iPhone was game-changing." Google, seemingly, agrees. In May 2016, it announced Home, a voice-enabled wireless speaker in the mould of Amazon Echo. "Once you have a device that's always listening, you have a different relationship with speech," says O'Reilly.
In October 2009, Apple launched in-app purchases for the App Store. The software industry hasn't looked back. In the second half of 2013 alone, Candy Crush Saga made $1.04 billion from microtransactions. More recently, Pokémon GO, Niantic's runaway-success game, made $35 million from in-app purchases in two weeks. According to analysts IDC, revenue from mobile apps, not including advertising, was around $34.2 billion in 2015.
For bots, the opportunity could be even greater. "Bots have emerged as a high-potential channel of distribution for mobile services," says Guo. Not only do messaging apps have a captive audience, the cost of developing bots is lower than for apps. "The progression from trivial to sophisticated is going to happen faster," says Underwood. "App developers have been able to learn from the introduction of prior interfaces because it wasn't long ago that mobile apps came on the scene. It took a few years in mobile. With bots I think it will happen in half the time."
Libin, one of the bot industry's leading investors, has no doubts about its transformative potential. "There are going to be 100 million bots. It's going to be similar to the app gold-rush, but magnified," he says. As with apps, the vast majority of bots will be pointless, he argues. "But the few hundred that are actually really good are going to be world-changing."
This article was originally published by WIRED UK
