Navid Hadzaad is the CEO and founder of a startup called GoButler. You may not recognize the name. But if you follow the ever-changing world of modern technology, you'll recognize the plan. GoButler wants to connect people with businesses through mobile messages. You know: texts and things that look like texts.
In recent months, this rather simple idea for changing the way customers interact with companies has pervaded the tech industry. GoButler and its eponymous smartphone app use mobile messages to provide a kind of digital assistant---with a few mobile messages, you can book a plane flight, buy some flowers, and order a pizza, among other things. You can do much the same through services like Operator and Magic. Facebook is building a similar system inside its Facebook Messenger app. Google is building one too, according to The Wall Street Journal. And this week, at a conference in Munich, the Facebook-owned mobile messaging giant WhatsApp announced that it too will expand into "commercial messaging," eying connections between consumers and businesses as a prime way of making money.
Approaches may differ in the details. Some services are paying a team of customer support types to read all incoming messages and provide a very human bridge to the right businesses. Others hope to automate the process, building "bots" or other artificially intelligent systems that can process your requests without help from human workers (or only some help). In a few cases, companies are combining these two approaches in the hope that one will fuel the other. If these companies collect enough data on how human workers behave, the thinking goes, they can teach machines to do the same stuff.
GoButler is a representative example of this last strategy. It runs a team of customer support types that can easily parse requests from users, but it's also building AI that over time is learning to make sense of an ever-widening range of human wishes. The difference is that Navid Hadzaad is wonderfully candid about this would-be market. Eating breakfast at a Munich hotel the day after the big announcement from WhatsApp, he acknowledges that all these companies are pretty much trying to duplicate the success of WeChat, a messaging app that has so successfully connected people with business across China. But he makes no bones about how difficult this will be---in part because of technological challenges, in part because its unclear if people outside of China really want messaging to replace apps and the web.
"I don't think anyone has evidence that this is what people want," says Hadzaad, who founded GoButler in Berlin but has since moved the company to New York. "But messaging has already had an impact on social [networking]. And I believe it will have a similar impact on business."
Julie Ask, an analyst with tech research firm Forrester, believes in the promise of messaging, echoing many other pundits. She argues that WeChat has provided "the playbook." And perhaps it has. In China, with help from simple bots and other tools, it's already a viable way of hailing a taxi, ordering food, and buying movie tickets. The rub is that this "universal mobile interface" evolved in a market that didn't necessarily have dominant alternatives. That's good news for WhatsApp, which is enormously popular in developing countries that are still coming onto the Internet. But the prospects are different for messaging apps in the US and similarly developed markets. In these places, viable alternatives already exist. You can easily hail a car from Uber and Lyft's existing apps. You can order food from GrubHub. Why would that need to change?
The other wild card here is AI. People may like the idea of a digital concierge that can handle all their basic daily tasks. But driving such a service with human operators isn't exactly that most economically viable idea---at least not in the long term. Magic is now charging $100 an hour for its human-driven service. But AI can change the economics. That's what Facebook and GoButler and others are trying to do. And apparently, it's what Google is eyeing as well.
In some cases, companies are exploring fairly simple bots that only respond to particular commands or code. But this too seems unlikely to work. That's not how people use their devices nowadays---at least not here in the US. As Hadzaad says, what's really needed is AI that can understand and respond to natural language. That's what GoButler is striving for---as is Facebook.
At the moment, these companies are collecting data on how human workers handle requests, and then they're feeding this data into deep neural networks, networks of hardware and software that can learn tasks by analyzing large amounts of information. These neural nets can learn to identify animals by analyzing large numbers of animal photos. They can learn to recognize spoken words by analyzing voice commands. And they can learn to mimic customer support by analyzing how these humans respond to requests---at least in theory.
But as Hadzaad will tell you, this kind of mimicry isn't easy. Getting it right will take time. GoButler is preparing to launch a system based on this kind of deep learning, but it will focus only on one particular task: booking plane flights. Other tasks will come later. "We won't be doing everything," Hadzaad says. "We'll be automated, but what people are allowed to do will be more structured." In other words, we've reached the age of commercial messaging. But we'll have to wait and see how smart it really gets.
