Home >News >India >Uber’s obsession is automation, but it can't work where safety paramount

NEW DELHI: With more startups selling themselves on the fact that they’re tech enabled, customer support is quickly being driven towards more automated systems. Mint spoke to Vidya Duthaluru, who has been leading Uber’s Customer Obsession Engineering team in India for two years, and was recently elevated to Director, Customer Obsession Charter for its global operations, about how the company approaches customer support. This is the team that builds all of Uber’s customer support platforms, for both drivers and riders.

Edited excerpts:

Uber’s approach to customer support seems to be to build a completely automated system. Is that how you think about it?

Our approach is to take care of specific issues in the most efficient manner possible. We’re investing right not, into the ability to proactively understand what you might be contacting Uber for. Can I look at certain signals in the ride and decide why a customer is contacting Uber, so we can deal with that efficiently.

Obviously automation is the most efficient way to solve the problem. But there are other personal issues where you absolutely need a physical hand. For instance, safety issues are never automated.

How does Uber arrive at the carousel of five or six options that a customer sees when contacting customer support?

Today the options you see are somewhat dynamic, but not fully. It’s reasonably static. However, we’re actively working on making technology that will allow for a far more dynamic set of options. We come up with these options based on your interactions with Uber. For example, if you’ve just taken a ride, there are a set of things you are more likely to contact us for. We apply machine learning to show the ones that will be featured. We can’t make any predictions if you haven’t taken a ride in say the last 50 days, and then the options will be more generic.

Since any automated system will lack empathy. How do you account for that in a system like the one on Uber?

It’s really about the type of issue you’re looking at. We have processes in place where we are able to give priority to your request, able to find out the request to the driver, we also have Partner Seva Kendras where drivers can have human interactions. We try to combine those processes along with tech to solve issues in a more efficient manner. Sometimes process and operations teams come together to provide that empathetic angle. In certain issues, empathy isn’t super important, which we look to automate. That’s why we can never automate any kind of safety issue.

When we talk about the use of artificial intelligence (AI) or machine learning (ML), there’s a certain learning aspect expected. Could you explain the signals that these algorithms are getting in Uber’s system, other than the obvious ones like user feedback, button taps etc?

There are several signals. Like the GPS locations during the trip, what was the initial and final fare, initial estimated time or arrival (ETA) and total time take etc. We can use a lot of these signals at an aggregated level to check what might have gone wrong. These are certain signals that we typically look at to predict when a trip may not have gone wrong.

Do you think automated systems are ready to replace human support systems?

I’ve worked in this space for a while and I would say that we will get there, but we aren’t all the way there. We are definitely at hybrid systems there, where AI and ML can work well if you have trained them with the right level of data. Having said that, there have to be certain guardrails.

Subscribe to Mint Newsletters
* Enter a valid email
* Thank you for subscribing to our newsletter.

Never miss a story! Stay connected and informed with Mint. Download our App Now!!

Edit Profile
My ReadsRedeem a Gift CardLogout