Google One’s Launch Shows the Hurdles of Automation For Customer Service


Companies looking to add automation to their customer service workflows should pay attention to the case of Google One, which the tech giant revealed on Monday. It's a paid subscription that promises human support for any problems people have with Google’s consumer products.

This may seem a bit odd, considering just last week Google showed off an automated system called Duplex that uses machine learning to power a computerized agent to talk on a customer’s behalf to perform certain tasks like scheduling a haircut or inquiring about restaurant reservations. After all, if Google can create a digital assistant to speak for its customers, why can’t it use that technology to speak to its customers?

The answer illuminates one of the key realities of customer service: there’s still no substitute for humans for tackling difficult edge cases. ISG research shows that consumers, especially millennials, are happy interacting with digital systems for service tasks, in part because those platforms can provide faster resolution. But consumers also are less patient. Google productivity services like Gmail, Docs and Drive are critical to the way many people live and work, but the company to date has made it difficult for consumers to acquire official, personalized help.

Google’s philosophy has focused on creating the most reliable services possible, then providing customers with the self-service tools necessary to solve their own problems. That’s a useful approach when working at Google’s scale, but it lets some problems slip through the cracks. Until now, consumer users of Google services have had to rely on whatever they could scrounge up on technical support forums and the web at large. The tech titan’s overall focus on service quality and self-service support seems unlikely to change, however.

Adding an extra dose of humanity to its customer support process also could help Google improve its automation going forward. It could use support requests from paying customers to create training data for more advanced support bots in the future, for example. What’s more, these requests allow the company to acquire more data about its products’ performance with its most committed users. Product leaders can use that information to help prioritize new features and fixes.

Google One’s example is an important one for enterprises looking to transform their customer support practices with automation. If one of today’s market leaders in creating conversational artificial intelligence can’t produce an automated system that serves all its customers’ needs, that shows how nearly impossible the task is. Instead, modern customer support calls for the use of a multi-modal approach: automate what you can to provide better customer experiences, then provide highly skilled humans to help with the rest. That’s especially important for paying customers, who expect service providers to help with what they’ve purchased.