You'll notice something quirky if you ever visit the offices of virtual assistant startup Ozlo: Stuck on the leaves and trunks and needles of the many ferns and bonsais and cacti throughout the office are pairs of googly eyes.
The decor is an allusion to a 2008 Saturday Night Live skit where a man frightened of plants (Christopher Walken) tells viewers as a gardening tip to place eyes on their greenery. "Normally plants don't have eyes, so it's very hard for me to trust them," Walken explains.
The challenge of making something alien feel friendly is one familiar to Google, Apple, Amazon or any company offering virtual assistant features--and to this 30-person startup with offices in Palo Alto and Seattle.
The solution for Ozlo has been to design its artificially intelligent assistant as a childlike mammalian character with an easy-to-pronounce name (Ozlo) that can learn from users and contextualize his own limitations when he encounters them. (Ozlo is gendered male.)
"We think of him as kind of young," and by virtue of his youth a little helpless and still learning, Ozlo CEO and cofounder Charles Jolley says on a sunny September afternoon at the startup's Palo Alto office. Jolley is former head of platform at Facebook and ran Facebook for Android.
Standing in front of a wall plastered with sticky notes describing ideas for Ozlo's characteristics and images of other virtual assistant avatars, he adds, "We had a version where he was in diapers, but that was too young."
With a $14 million round of funding from Greylock Partners, AME Cloud Ventures' Jerry Yang and other investors, the potty-trained version of Ozlo came out of private beta this week and is available on iOS and the web. An Android version is slated for release this year.
Aside from his bulbous eyes, his balance of seeking to learn while acknowledging limitations stands out as his most prominent feature. An example of how Ozlo displays this quality is an ability to intuit mistakes a user might make, sussed out through his understanding of different words and terms.
Jolley demonstrates this on his phone, typing into the Ozlo app, "Mexican called del medio on California Ave." Ozlo determines that the user is referring to a restaurant and knows the phone is in Palo Alto, so he centers his search there. He knows that there is a Cuban restaurant nearby with "Del Medio" in its name, and he knows that Mexican and Cuban food can both be classified under the larger penumbra of Latin cuisine.
"I do not know which restaurants serve Mexican, but La Bodeguita Del Medio is a Latin restaurant on S California Ave," Ozlo responds, as if to say, "I think I see what you're getting at and here is what you might actually be looking for."
Sometimes Ozlo's intuition gives out. When I asked the app when I should leave Ozlo's office to catch a train back to San Francisco, Ozlo told me where the nearest train stop was without any follow-up questions or explanation that he didn't know the train schedule but could tell me where to find the stop.
Ozlo also avoids making judgment calls in his responses. Ask him for "the best restaurant nearby," and he'll respond with a carousel of what he describes as "many good restaurants nearby."
The idea here is to "be the librarian and not the book," says Ozlo CTO and cofounder Mike Hanson, former principal engineer at Mozilla. Through what he describes as "mindful design," Ozlo indicates "here's why we're making this claim," appending snippets of reviews and labels such as whether the user has designated the restaurant as a favorite spot.
Why all the stuff about food? Because Ozlo's skills are still a bit limited. For now he can answer queries related to restaurants, coffee shops, weather and recent headlines. He can also show showtimes for movies in theaters and book tickets for users, order you an Uber and arrange for a delivery, Jolley says.
Next up for features: movie recommendations incorporating information like plot, reviews, cast and where the film can be streamed; and restaurant reservation booking. Voice features are not being released at this time.
These limitations are strengths, according to Ozlo's founders. Because the company is still a relatively small startup, it has the freedom to hone a smaller slate of skills rather than try to do everything at once. "Google could never launch something this narrow," says Jolley.
And in a sea of virtual assistant options like Hound, Viv, Facebook's M, Alexa and Siri--Grand View Research predicts the size of the global intelligent virtual assistant market will reach $12.28 billion by 2024--he sees a Google-sized opportunity.
"I think the opportunity is that big," he says. "We're kind of at a similar place that we were in 1999."
What he means is that by working to synthesize information from multiple sources and present it with clarity and context, he thinks Ozlo is doing something as significant as indexing the web was, nearly 20 years ago. And he and Hanson believe Ozlo's skills will snowball. Dinner and a movie tonight--and with time, everything else.
"I mean, we want him to know everything," says Jolley.