Virtual assistants powered by artificial intelligence are a staple of popular culture, from HAL 9000 in the Space Odyssey series to Samantha in 2013 film Her .
But these assistants are no longer science fiction. There’s Apple’s Siri, Microsoft’s Cortana, Google Now and the recently-unveiled Facebook M . Plus, there’s Slackbot, the “assistant, notepad and programmable bot” whose potential is one reason workplace communications startup Slack is valued at $2.8bn by investors .
Slack started as an internal tool for Tiny Speck, the games company launched by Stewart Butterfield, who’d previously co-founded photo-sharing service Flickr before selling it to Yahoo. Tiny Speck’s first game, Glitch, wasn’t a hit, but the internet messaging software used by its developers had potential.
One pivot and $340m of funding later, Slack is the talk of the technology industry, with more than 1.1 million people using its apps and website to message their colleagues, as well as its resident chatbot.
Slackbot is first encountered when it helps you fill out your profile on Slack, with basic uses including the ability to store any messages you send it (the “notepad” in the description above) and to be programmed with a range of automatic responses to lightly troll colleagues.
Slackbot and similar chatbots’ future potential, though, lies in the ability for developers to programme it to perform more useful tasks.
The New York Times has created its own Slackbot called Blossom that recommends which stories are worth posting to Facebook when asked, following a previous experimental Electionbot that posted election results in a relevant Slack channel for journalists.
In the future, Slack’s ability to hook into a range of other online services could see such bots become valued colleagues: pinging messages back and forth to book work trips, research relevant topics; run recruitment campaigns and much more.
“Slackbot is nothing: we’re right at the beginning of this. Over the next several decades, new applications of computing technology in people’s lives will increase productivity as dramatically as computers did in the 1960s and 70s. And it will be about making us more efficient rather than taking our jobs,” says Butterfield.
Talking about the future of AI assistants at work involves first looking back, with Butterfield pointing to the advances in productivity in the early years of computers making their way into the workplace.
“Using computers to do accounting versus people doing double-entry was huge, because it was something that computers do very well – doing mathematics and logic operations very quickly – which humans don’t do so well. We saw huge increases in productivity,” he says.
Related: Why Slack is worth $1bn: it's trying to change how we work
“But then we hit that period of the 90s to the early 2000s, when the easy gains were already taken advantage of. And so we saw more questionable applications: Rather than an executive dictating a letter to a pro typist who could type at 120 words-per-minute, we’d get the CEO fussing around with Microsoft Word taking an hour to format a single paragraph right.”
Butterfield, his investors and a number of his tech-industry peers hope we’re on the verge of another big leap forward in productivity, fuelled this time by inventive machine-learning technology and personable, programmable bots.
Is that a promise about them not taking our jobs? “I am sceptical about machine intelligence’s ability to replace what humans do. You do see phone systems that increasingly replace people who work in call centres, and as they get better at voice recognition and speech-to-text, they will be able to handle customer support requests, but it will be mostly simple things,” says Butterfield.
He thinks that the importance of human interaction will often prevail, citing the example of experiments to replace staff at fast-food outlets with kiosks and software.
“The menu is pretty short and there’s not a lot of complexity, so machines should be able to do it, but what they found is that people don’t really like that compared to ordering from a human,” he says. “Face-to-face interaction makes that process a lot better, so there’s still a lot of times when people are going to be required to talk to other people.”
Facebook Twitter Pinterest Slack chief executive Stewart Butterfield.
Instead, his vision of what Slackbot and similar assistants will do for people is partly about administrative tasks: asking everyone on a team what they’re working on every morning, and then distributing that information, as one example. Or making sense of the (he estimates) 80% of our work emails that come from machines rather than humans.
“A company like us gets 60k to 80k messages a day: every time someone tweets at us, customer support tickets, when the mobile app crashes or a new bug is filed, when someone makes a purchase… and that is all filed into Slack and becomes searchable,” he says. “We are going to see more and more of that.”
Butterfield sees Slackbot as playing a useful role in this, in contrast to what he sees as some of the more-hyped competition in the world of bots: in a recent Wall Street Journal interview he described Apple’s Siri as “fucking idiotic ... nearly useless” for example.
Don’t take this as an argument for Slack as an email killer, as some overexcited press coverage has. In fact, one of its recent announcements focused on helping people bring their emails into their company’s Slack channels . Butterfield freely admits to spending five hours a day on email, albeit only for people outside his company.
“Email is the lowest common denominator of communication, and we’ll continue to have it for decades. But many teams who use Slack either reduce or eliminate their internal use of email. I would never think of emailing somebody else within the company,” he says.
This is what Slack is pitching as its “big benefit”: the ability for companies to reduce internal emails in favour of messaging within Slack, gradually building up a corporate archive that is searchable and – ultimately – made even more useful through Slackbot and machine-learning.
“Slack messages are typically addressed to channels rather than individuals. This idea that a message could be addressed to a channel that exists before you get there and after you leave is the fundamental difference to email,” says Butterfield.
It’s a corporate or organisational memory, which is a concept that another highly-valued work-focused tech firm – Evernote – has made a key part of its pitch to businesses too.
“Your co-workers have all this information, but the company has forgotten what it knows. It knows it’s there, but not how to get it,” Evernote’s then-chief executive, Phil Libin, told the Guardian in 2012.
The difference is that while Evernote talks about notes and collaboration in a “digital workspace”, Slack is focused on messaging. And it is “messaging”: don’t call it chat within earshot of Butterfield.
“We’re forbidden from saying chat internally. It makes it sound trivial. No important conversations are ‘chats’ unless you’re getting fired: that’s the only time your boss will invite you for a ‘chat’ that’s important,” he says.
“We are also forbidden from using the word ‘collaboration’. There’s a 30-year history of broken promises around computers’ ability to help you collaborate ...”
Related: Evernote boss talks zombie emails, Apple Watch and sci-fi collaboration
It’s hard not to read that as a criticism of Microsoft in particular: the company that has dominated workplace technology for the last 30 years, but which is seeing new competitors to its software emerging from all angles, chipping away at the idea of the do-everything behemoth.
How does Slack avoid becoming the next Microsoft in a negative sense, adding so many features that it becomes unwieldy, and open to similar attacks from smaller, nimbler startups?
“We still want to become the behemoth, but in a different way: we can only do a very small number of things well. Microsoft used to do everything: people would buy most of their software from the company, and it all worked together,” says Butterfield.
“Now, for every product category that Microsoft once dominated, there are a dozen good competitors. Businesses often end up buying from three or four dozen vendors, and the software is cheaper, more powerful and easier to deploy and manage. Across the board everything is better. Except that none of this software works together.”
Butterfield hopes that messaging is the logical place for “this to all come together”, meaning his company can focus, not on trying to do everything like old-school Microsoft, but on integrating with all these different services and (with the help of Slackbot) making them better.
“If we make them just a little bit better, we can provide a huge amount of value to our customers, and people will love us and want to use us forever,” he says. “But let’s see where we are in 30 years’ time …”