Last year in December, Booking.com, one of the world's largest online travel companies, expanded the pilot version of its new service and support chatbot, the Booking Assistant, worldwide. The company had first announced the message-based customer service platform in mid-2016. With a team of more than 2,000 technologists globally, Booking.com has been experimenting with emerging technologies like artificial intelligence (AI), machine learning (ML), and natural language processing (NLP) for a while. In September 2017, it acquired an Israel-based AI company that had developed a virtual travel agent. In an interview, Booking.com’s Vice President of Customer Service, James Waters, shared why the company needs an AI-powered chatbot and how the customer service will evolve and transform in the future. Edited excerpts.

Interviewed by Moulishree Srivastava

/ / When did Booking.com start tapping technology to drive its customer support, and how that has evolved over the years?

We started working on building a bot about 18 months ago which is now the Booking Assistant. The traditional channels to connect with customer support and customer care are email and phone. Email is too punctuated---you send an email, you go away and get a reply many hours later. A phone call is too invasive. Suppose you are standing at a bus stop. You can open a text conversation, you can carry that on. If you need to pause it for few minutes to get on the bus and buy a ticket, you can do that. It gives more freedom to the guests. That was our first belief.

Our second belief was that it doesn't make sense really for us to open up a chat functionality that doesn't serve the purposes that email and phone are serving. What we said was, we need to have a chat function to support the communication that our customers are engaging in with us today. What it may also do is open up a channel for a new kind of support, a new kind of communication later, especially with Booking.com expanding its offering from just accommodation to attractions, and some transport offerings.

I think messaging and chat are much more powerful as channels for us to be in contact with guests during their stay than phone and email. Again, you don't want to be on international roaming charges on your phone while you are on holiday. And you don't want to be scrolling down your hundreds of work emails to find the one from Booking, so the chat function is still the better way to have a consistent relationship with guests. What we have been doing is making sure that the Booking Assistant can support guests with the most commonly asked questions and also get in new types of questions, which is an opportunity for us to serve our guests better.  

/ / How has Booking Assistant changed the way Booking.com handles the customer support services?

Our Booking Assistant combines machine learning and AI and has the provision for our guests to connect with and talk to a human being on the other end if they need to. The idea is to solve the issue that the customer has at that point in time. So if the customer has something that's mundane and functional, it's fantastic if technology can solve that problem instantly. Equally, if they have a more complex and unusual situation where they need to connect and talk to a human being, we allow them to do that. Thirty percent of the questions that are asked to our assistant, we are able to solve within five minutes with our technology.

/ / How has the customer support service evolved over the past few years?

Over the last few years, we have seen two strong trends emerging. On the one side, there is a push for personalization, for the human element, and on the other side, the push for more speed and ease. What that actually means is that if you want to provide a good service experience in the travel industry, you can’t have an option between efficiency and personalization. You need to be able to offer both depending on what the guests want. So if a guest has been on a plane for 15 hours and turns up to check in, then he would not want to stand in a line. If he can press two buttons and get in his room, he is going to do that. But the next day, he might want to have a long conversation with the local expert on the best things to eat and the best things to see.  We try to provide customer support in a way that allows for both, and that is where our Booking Assistant comes in.

/ / Does automation impact or change how people work within the company?

Sure, it has an impact on what people do, but we believe it is not so much that it replaces people, it pivots a little bit what people do. The idea is that the more common and repetitive things would get done through technology, and our people would pick up truly unique human cases like if your child were to fall sick or your flight gets canceled because a volcano erupted in Bali. So it makes it even more important that our people are strong on problem-solving and empathy.

/ / How are you using technology to improve your customer service experience?

If you look at how our Booking Assistant works today, when you first open it, depending upon where you are, we offer you some specific suggestions on things that you might like to do. Alternately, you can type in a question, and our system will attempt to understand what the question is and what the appropriate reply should be and if it is comfortable doing that, it is going to send you a text reply. But if it is not quite sure about the reply, we get our customer service agents involved. We track what our agents do, and how and what replies they give. So one of the things we try to make sure as the key component of what we do in customer service now, irrespective of whether you interact with the Booking Assistant or an agent on the phone, is understanding who you are as a customer and what you are likely to need from us before we actually talk to you.

In all this, we make sure there is a learning loop. Whatever it is that our people do, it is captured in a structured way and fed back into the Booking Assistant. So over time, bit by bit, we are able to serve more and more things instantly with technology.

What we also hope to do is make sure that all the things that we learn through the assistant are applied to the other channels that we serve as well. So if we get smarter in serving our guests through the assistant in a messaging environment, that learning can flow over in how do we do it on email.

/ / What are the technologies that you are using? Are you planning any new initiatives?

We are gradually moving more and more to the cloud which allows adopting machine learning at scale. We have made some investments in technology to be able to take sound or voice and turn it into text.

We have also invested in building an internal system for prioritization. We have many customers calling us at any given time. Some requests are urgent, some are less so. Some customers we know speak four different languages, and we can answer them in any of those languages. While some speak only one language, and we have to respond to them only in that language. How do we decide which customer to serve first and which customer to route to which agent - that's the technology that we have built in-house.  Our system also looks at who the customer is, when did they check in, what do we think that they may want. If you are checking in this afternoon, you need an answer now...if you are checking in on new year's day which is 6 months away, you probably don't mind waiting a little longer.

/ / What are the challenges that you face while trying to adopt new technologies to serve customers better?

The illogical nature of how human beings think and talk makes interactions very challenging.  Take, for example, a guest who may have multiple bookings with us. The guest may ask five questions about one booking and then switch over and ask a question about a completely different reservation. That's a tricky challenge. People switch context at random and it’s hard for technology to understand that.

The second big challenge that we see is that we are super strict about the quality of the experience that our customers go through with the assistant. It is a tussle between how quickly we learn and the quality of experience. The more data we gather and the more people we get into the assistant, the faster that technology will learn. But if we open the gates too wide for too many people too soon, the assistant would not be able to deliver quality experience. So we really have to balance the speed of learning with the access to enough number of customers.

/ / What are some of the trends that you see emerging in customer service in the next few years?

Technology does get smarter and it will be able to remove some of the high repetition work that is done by people today. I think that trend has several more years to play out. Messaging through voice—that’s the trend that I see emerging in the near future.

I think as we move from supporting just the accommodation to supporting the entire trip, there will be a slight pivot to our customer service. It will not only be about answering questions or problems but more and more about establishing a relationship and providing a support and information channel for travelers. That trend has begun and it will expand as we go forward.