advertisement
advertisement

Slack Invests In A Part-Human AI For Business

The email-based virtual assistant Clara has secured $7.3 million in funding to hook up with Slack, and more.

The best, and only, personal assistant I’ve ever had was Clara. She studied my schedule closer than I did, knew hours I liked keeping free from calls, and referred all meetings to my favorite local coffee shop. I’d just CC her on any email, and she’d take care of the rest.

advertisement

Clara, of course, wasn’t real. She was built by the startup Clara Labs, using a cross of AI and real people to handle scheduling for me. Now, Clara is growing up. The company has recently secured a Series A round of funding worth $7.3 million, with contributions from ex-Dropbox investor Lan Xuezhao’s Basis Set Ventures, Sequoia, First Round, and, perhaps most interestingly, Slack.

[Image: Clara Labs]
As Clara Labs CEO Maran Nelson put it to me, she believes her company has reached a point where it has both proven there’s a market for subscription virtual assistants and trained its AI technology enough to handle most of the heavy lifting, rather than rely on people. In other words, the market-tested product can actually scale.

When I last tried Clara, she was mostly people who were hired to respond to emails and schedule. This work was never intended to be waste. Instead, Clara Labs was gathering data to train its own machine to have one perfect conversation that might play out a million different ways: scheduling a meeting.

[Image: Clara Labs]
“You don’t just start with a bot from day one. You start very much manually. So the early early days, [my cofounder] Michael and I were manually scheduling meetings. Then I got my sisters to do it. Then we were working with other people to be Clara on the back end of the service,” says Nelson. “It was a means to bootstrap our way into the data we needed.”

Since, Clara has continued to get smarter, first suggesting responses to human emailers, and now, I’m told, manages to take most of those conversations over herself. While Nelson wouldn’t commit to a percentage of responses handled by Clara, she says that people are only flagged in when Clara doesn’t have a high certainty of offering a perfect response.

[Image: Clara Labs]
It’s a balance that the company refers to as “human in the loop”—in which humans are a safety valve to ensure quality, and to continually train the machines to get better in a case-by-case basis. Especially for a small startup, without the resources of a Google to throw at its smart algorithms, this design is allowing Clara Labs to go to market quickly.

advertisement

“If you think about the challenges of machine learning, the fundamental challenge is you can’t ship something to the user that doesn’t work reliably. You need to protect the user from the mispredictions, or the low confidence cases. These systems are in their infancy,” says Nelson. “We want people to understand what it is to have a hybrid system, and many systems work like this. With pilots, most of the time in the air, the plane itself is flying you. But the pilot is there just in case.”

[Image: Clara Labs]
Now, with the conversation and scheduling bot largely built and seemingly scalable, Clara Labs is looking to fit Clara into more spots than just your email, to help specialists like recruiters and salespeople leverage the service more effectively. The company will soon take the leap from personal to professional accounts–which might explain why Slack chose to invest. Clara will soon be sold in team- and business-level packages, much like Slack is today. And with more users plugged into the service, you can imagine being able to @clara in a Slack room to schedule a meeting. Rather than needing some complicated, natural language conversation to make that happen, Clara will simply look at everyone’s schedules, account for their individual preferences, and put a working meeting time on the books.

“Clara is on both sides of calendars and makes instantaneous, magical things happen,” says Nelson of this new model–even if that model is, at some points, as much of an illusion of AI automation as it is true code magic.

About the author

Mark Wilson is a senior writer at Fast Company. He started Philanthroper.com, a simple way to give back every day.

More