Demystifying AI for Airlines

by 
David Kaplan
Tuesday, January 21, 2020
 • 
9
 min read

“Whenever you see delays, cancellations or re-routing through to another city, often when you see a lot of passengers not happy for whatever reason. That's an area where airlines can do a good job, a better job, in deploying machine learning and understanding,” says Experience The Skies’ Larry Leung. 

It’s one thing for a commercial airline to talk about its retail digital transformation. After all, announcements of new artificial intelligence programs and app upgrades are made by carriers’ executives practically every week.  

It’s another thing to prove what digital transformation actually means in practice. 

Last week, we reviewed Delta Air Lines’ strategy for its rollout of personalized digital billboard messaging platform Parallel Reality with Larry Leung, founder and director of Research and Strategy at Toronto-based aviation/travel tech consultancy Experience The Skies.  

In the case of Parallel Reality, which will debut in Detroit Metropolitan Airport in mid-2020, Leung says utility should be the foundation of the technology’s use. Developed by engineering/design startup Misapplied Sciences, Parallel Reality uses digital signage in public places, such as an airport or in the corridors of a sports venue, to display specific image messages to individual passersby even within a crowd. 

In this wider conversation, Leung shares his views of how airlines should approach more general use cases involving AI/machine learning, loyalty/rewards programs, and revenue management/dynamic pricing. 

*** This interview was done as part of Kambr Media’s role as an Official Media Partner of the Aviation Festival Americas (May 12-13, Miami) ***

Kambr Media: How did you start Experience the Skies and what kind of clients do you consult with? 

Larry Leung: I started Experience the Skies about five years ago. I had a background in IT and customer experience through Price Waterhouse and Coopers and TD bank group. I was already doing a lot of application IT security and was focused on integrating helping businesses figure out what those applications mean to the customer. 

In addition, I really loved aviation as a kid. I remember going to the airport with my father when I was maybe six or seven and just watching planes take off and land. Since then, I’ve had a fascination with flying, and with travel in general. I realized that I wanted to get into a space where I can help customers through the tech experiences I already had already been working on at consulting firms. Everybody travels and after speaking to a number of people, I realized that there is a gap in integration, specifically when it comes to issues involving communication. 

When I started the company, I asked myself: What does it mean to bring a better customer experience? What are the marketing, the technology integration and loyalty program and communication components that bring the travel journey a better thing for everybody? 

At the end of the day, travel is a storytelling exercise: you go on any trip, business or leisure. At the end of the trip, you want to tell someone about it. "Oh, I experienced something amazing and I want to go back again,” or “That didn't work out for me, so if you encounter this type of problem, this is what you need to do." Everything's about storytelling. So I want to build that as part of travel journey. 

Where do you think airlines are in terms of their “digital transformation” at this moment? 

It's definitely uneven. The term “digital transformation” can be very confusing for a lot of people, because there is no true definition to it. It depends on what the companies are doing to manage the digital platforms and how to really use data and really what is the digital strategy. 

With that in mind, I think it's disjointed working on digital transformation between airlines and their airline alliances. We already see that. We see people use digital transformation for anything from the beginning, which is where they engage a customer to their service and products, and to the onboarding process, where they acquire you as a customer. From there, it’s about letting you go through the motion of using the products and services. 

If a company or airline or a hotel were to say, "We are going through a digital transformation process," that means they should look into every touchpoint possible to see whether or not there could be interactions between the company and the customer.  

Unfortunately, from what we see being posted by various companies, see a lot of times we see tech that still involves a lot of manual processing in place, which could have been digitalized. That's when every digital process seems to break down for some reason and that there are more confusion and more frustration. 

In terms of the digital transformation progress in terms of airline retailing, what’s your sense of the importance of IATA's NDC program?  

At the end of the day, the mission is to engage the customer so that the customer would be buying a product and services – or the upsold products – afterwards. Offering the clearest information about those products is very important. What IATA's NDC program is trying to do is to create a better communication format for displaying airlines’ products and services. 

The idea is great, because before there are many pieces of information, and it's all haphazardly put together. If a travel agent really wants more information, it may be very difficult to get those details directly through a system. As a result, they would not be able to do the best job possible to provide information to the customer. 

So far, IATA hasn't really pushed NDC to all airlines, unlike e-ticketing back in the day. A lot of the components are flexible and they're not mandatory. With that in mind, we know that the uptake hasn't been high, because it's a new platform. You need to hire a lot more people to get it set up. Also, there is not enough instant return on investment for the airlines.  

In the end, they are sharing information to other sites, whether or not it's through a third party intermediary helping them push the end process NDC or the end travel agents. 

To some degree they're sharing information that they kind of want to. But they also don't want to, because they want agents and individuals to buy directly through them. So with that dilemma. Some of the airlines may not feel like they want to push forward as fast as they want to, IATA wants them to.  

Do I see specifically whether or not it's useful? I absolutely think that NDC would be very useful for the end user, whether or not it's corporate travel agents or end users like me and you at the end of the day, because we want to be able to get the best forms of information available to make an informed choice of the products and services that we want to have. 

What are some examples of the product confusion NDC can address? 

I often use the word “basic economy.” In the U.S., basic economy fares are very popular. In Europe, they are named many different types of terms, mostly proprietary terms so that people would be able to differentiate them.  

For example, on United Airlines’ basic economy, you can't bring in a carry on, on American and Delta you could. As for the customer, they 

That happens extremely often. And it's unfortunate, because if NDC is working the way it should be, then when the customer is buying the fare, they should get what they’re expecting. For example, code-sharing has been in place for more than two decades. And yet many people still don't really fully understand how code-sharing works and how fees could change because of code-sharing. That's where NDC could continue pushing forward.  

And another area of confusion and frustration with fliers is the changes that are happening in mileage points, loyalty rewards programs as airlines shift to dynamic pricing and the changing values of miles. I know you've had a lot of thoughts that you've shared in the past on that topic. 

There has always been an imbalance of what an award really means to an airline. When I was at PWC, I worked on audits and Air Canada was one of them. Just look through there any airline's annual reports, you can see that there is a liability related to miles that are sitting on their end, waiting for redemption. A lot of airlines want those miles gone and not on their books, to some degree.  

But exactly how do you change the way the miles behave if you only use the fixed number, which has nothing to do with cost? Sometimes you lucked out and sometimes you're not allowing people to use those miles. Then the question is “What to do next?” Dynamic pricing. And it's definitely a not new concept. 

It gives back balance to some degree for airlines to say, “Okay, why wouldn't it be different if I only have one seat left to fly from say Toronto to London that I'm going to charge that at the highest possible price? Why wouldn't I do the same thing with miles? If the seat is available for a mile redemption, why would I charge the standard price when that's what I charged at the beginning?” 

I would actually see more airlines looking into this particular dynamic pricing model. I would like to see more airlines do away with the award charge to a large degree, which Delta has done. You actually don't see an award chart anywhere, it's all dynamically priced.  

Then what you would also see was the pricing go down. I have, during Christmas time, just before the 24th, 25th, 26th, all the air fares go high. But if you're traveling right on Christmas day or boxing day, prices can be low. There have been instances where people were able to find round trip tickets on miles to from eastern USA to Europe for 20,000 miles in economy. That would never have been available previously if a standard pricing is used. Because dynamic pricing was used, people were able to redeem at a lower price. 

Airlines really need to communicate the benefits of that, as opposed to just people think, "Oh, something's being taken away from me."  

What do you think of the role of AI and machine learning in airline operations, particularly in managing things like dynamic pricing in a revenue management context? 

Revenue management is the first place where they're using deep thinking, understanding inventory versus pricing and overbooking. When airlines have bigger networks, they have to use technology to guide them. Beyond just overselling, many airlines are looking into other KPIs such as customer satisfaction, in terms of how they could maximize revenue per passenger at a lower cost by using machine learning.  

Airlines can also do a better job in personalization. I don't think that has been done well. I'm sure you get emails from airlines every week of your chosen airline. They never say to you, "David, you've been searching my airline about going to Bali. This message is based on your views to go to Bali." Maybe you have never seen Bali ever show up in your email. And you don't really care because you're like, "Why would I go to somewhere in Ohio when I don't really care about Ohio quite yet?" 

I see it every week. That kind of engagement is definitely not personalization number one and number two. It stops you from engaging with the customer in a meaningful way, because you just don't feel like it's meant for you. From that angle, I see more AI being deployed and machine learning being deployed to understand what the customer wants. That's why it's more important for airlines to take control of the customer journey. Machine learning is not really just about buying the airline ticket, but the entire travel journey. And that's where the machine learning would be very useful. 

I review papers for a transportation research board every year. We’ve come across a lot of papers over the years on pricing, on whether or not university students are able to build models where they can look into pricing and revenue management so that they can manage overbooking better. Over the last say, three-, five years, I have seen those papers become more sophisticated. 

I would only imagine if some of those ideas were used by airlines, they would do an even better job than they do today on managing that part of the business. Is it really just revenue management? Is it really just about overbooking? I'd say no. The airline should think about, "Well how do I upsell you something? What exactly do you want? What can I do to entice you to buy a glass of wine?" 

Would some from New York City be willing to pay more versus someone from Kentucky? I don't know. But I feel like that could be data points to answer that question. After all, alcohol sale is a big thing for an airline. 

Another area where airlines may not do an amazing job, and you see more PR disasters because of that, would be disruption management. Whenever you see delays, cancellations or re-routing through to another city, often when you see a lot of passengers not happy for whatever reason. And I feel like that's an area where airlines can do a good job, a better job in deploying machine learning and understanding. 

They should know many markets. They know they're weather prone and yet receive the same stories over and over again on yearly basis of how passengers are being mistreated in one way or another. Partly it's education that the passengers simply do not understand what it means to be stranded and what are their rights. I go back to storytelling. Anyone can turn a bad situation into a good story. All that story depends on exactly what happens but it's a story. 

How do you change narrative and take control over that narrative and say, "I will do what I can to make it right for you." Using the machine learning programs to anticipate and prepare people so that they know what to expect means they won’t be as upset because the airline has been proactive about it. Machine learning can help airlines get ahead of those issues.