Travel insiders have long wondered when, or how, Amazon might flex its online muscle to follow other retail giants and non-travel brands into the sector.
Yet such speculation misses the fact of the company’s growing behind-the-scenes role in travel, mostly through Amazon Web Services, a subsidiary that provides on-demand cloud computing platforms and applications on a pay-as-you-go basis.
AWS’ importance to the larger company’s operations was clear in an earnings report this month when Amazon said revenue from its cloud unit beat analysts’ expectations by increasing 12% year-over-year in the second quarter to reach $22.1 billion. Of Amazon’s $7.7 billion in operating profit, 70% of it came from AWS.
As the largest cloud provider in a field that includes Google Cloud and Microsoft, Amazon – through AWS – has a significant toehold in the travel industry even without selling travel packages or booking rooms and flights, with cloud customers including Travelport and ATPCO.
Another is Trip.com, which reported significant improvement in its air ticket booking system and major cost savings after migrating more than 400 of its international business microservices to AWS.
Subscribe to our newsletter below
That partnership has led Trip.com to develop a joint innovation lab with AWS, focused on five topics: artificial intelligence, flight business, hotel business, international business and cloud technology.
“The innovation lab aims to make travel more accessible and provide the ultimate experience for travelers, using the latest technology, renowned experts and leading researchers from Trip.com and AWS,” Chen Zhang, executive vice president of Trip.com Group, said when the partnership was announced in April.
Meanwhile, Amazon said it has signed up travel brands including Lonely Planet and Ryanair for Amazon Bedrock, its service for building generative AI applications on Amazon’s cloud computing platform. In June, AWS built on that foundation by announcing a $100 million investment in a generative AI innovation center to work with customers on building and deploying AI solutions.
To gain a greater understanding of the company’s activities in the travel sector, we spoke with Massimo Morin, the global head of travel at AWS and a 25-year veteran in the travel industry. Our conversation has been edited for clarity and brevity.
I read that you said travel “is an industry that thrives in complexity.” Could you elaborate on what you meant?
If you fly, for example, the whole system that you are touching – for getting your flight ticket, for getting the reservation, going to the airport — you might not realize [how complex it is]. Many of these systems are still running on mainframes. … You have all these processes and procedures in place [with] legacy technology behind the scenes … but its original purpose is no longer there. So they trying to transform. The challenge that travel companies and airlines have is that expertise. The first thing is having people that know the new technology, but also having people that understand the old technology – why did we do it in that way? So that is the complexity that the industry is living on. If you think about it, you book your ticket, you go to the airport, you get the reservation, you get the taxi, the security, the check-in, the baggage, the gate, the plane, the pilot, the crew. You see there are a lot of bits and pieces that need to go right. And we don’t usually give them enough credit for the amazing job that they do.
Can you offer an example of how outdated technology affects the way companies interact with customers?
It starts with even very simple things [like] when you call an airline and spend countless minutes, hours, on the phone because they cannot actually transfer the call and they have to put you on hold. They cannot call you back, and they cannot understand who you are when you call and connect your phone number with your reservation. So they can say, “Hey, Derek is actually the person that is calling. Let me call him back because he’s traveling in a month, so he can [afford to] wait and I can [speak with] somebody else that is traveling tomorrow.”
They can’t do that [with the legacy technology], but with modern technology like Amazon Connect, they can do it today. The technology enables you to prioritize the queues. [And] you can have different types of support depending on your frequent flyer status or how much you spend and things like that.
How does that work?
Companies like Delta Airlines and Priceline are using our system called Amazon Connect. Amazon Connect is a contact service system in the cloud. You pay as you go, you scale it up and you scale it down depending on the needs that you have. You are able to activate the data that you have about your customer. You can do very simple things like a phone recording. You can do sentiment analysis [and] convert the voice into text. Now you understand if the customer is happy and what he’s talking about [and] we can enrich the customer profile. When is he calling? What is he calling about? What are the pain points that he was experiencing?
At the same time, you can offload a lot of these activities to a chatbot, for example, and therefore you enable self-serve. One of the challenges that the industry has today is that they do not have enough workforce, or they’re not trained enough, or [if] they are up to par with what they need to be … they’re a very scarce and expensive resource. So offloading to chatbots is actually a win-win because it enables self-serve [when] the new generation do not want to talk to anybody. And then you make the life of the agent [easier] because now they’re answering only the questions that are really requiring their attention.
You also have Amazon Bedrock, which is focused on helping companies build generative AI applications on your cloud computing platform. Tell us about that.
Now the gen AI is a bit transformational for the industry, because there is all this data you want to find meaning out of it and make it engaging with the purpose that the company has. The chatbot is most likely using models for customer engagement, conversation, communication in a much more natural way. So you type things like, “Where can I go this summer?” and the system replies.
Now to have this model to operate at scale, the [model needs] to be trained, right? So they need to be trained to test vast amounts of data. So you have to have two basic things. The first thing is you need a machine to run the training, so that when you ask a question, you get the [correct] answer. The second thing is that you want to have access to [the right] model. Not every company has all the expertise to build gen AI models. So we created a service, Bedrock, that enabled access to these models. You select the model that you want, and we have a variety of models that are coming directly from Amazon or from other providers we support in the AI labs. We work like Switzerland – we make accessible all these models.
You can train them with your specific data. This is key. One of the things that you don’t want is that when you train the model, [you don’t want] your data to go on the internet. You want the data to be private for yourself and for the specific use that you have. The second thing you want is when you train the model, it’s actually trained for your own purpose. So what we are seeing is that we need, because this is so new, we need to educate our customers on how to do this. So AWS keeps the data safe and secure, trains this model at scale, to complement the data that this model is trained on with the data specific for the business.
How is travel guidebook publisher Lonely Planet using the service?
They have no interest in using gen AI to write their guide. Why? Because the thing that differentiates Lonely Planet is that they have people at the destination. In Rome, you want somebody that lives in Rome to tell you about Rome. They have no interest in using a machine, and they have no interest in flying somebody from Wichita to go to Rome to do this thing. They have a point. They tell us, “Our differentiating value is our content, but now I need to create a mechanism for consuming this content in a much more accessible way.”
This is where generative AI comes in, because you can train the model, a model that speaks English, about what is happening in Rome. Now you can ask this model through a chatbot, right? What can I do in Rome? What could I see? Where could I go? If I want to go to the Colosseum, is there a timetable? When is it open? Those are very simple questions that come to your mind that you can find in a guide, but you don’t want to go through a book for it.
Lonely Planet is a 50-year-old-plus company. One of the interesting challenges they have is the new generation do not read books. They prefer to interact digitally. So how do they enable [these customers] to consume the company’s content in a much more efficient way? To me, this was a bit of an eye-opener into how the company that has the data, that has the content, is going to take [the] most advantage of generative AI and drive value to their own business.
Where do you see better technology taking travel businesses next?
All of these companies, they want to have more data. But the thing is to make sure that you have your house in order, break down the silos, connect all this data together and leverage it the way that you can. Now, when you want to start the journey, where do you start the journey into the cloud?
The first things that you can look at is analytics. It’s very low, low risk and high value. I have this data; I want to see what is happening. Where does my customer drop off in my purchase panel on my website? Are they purchasing all the product I have? If they don’t, why didn’t he buy the ticket? These are the kinds of things that you can easily do in AWS.
Another thing is what would make the marketing campaign more effective. Should I send a text message? Should I send an email? Should I send an app notification? I need to understand where the customer is at. These are the kinds of things that are very, very easy to implement. They are very, very low risk, but they help you understand better the customer and serve better the customer, and in reflection, improve your operation as well.