How artificial intelligence-powered apps are streamlining how we travel.
A record 28.5 million people will fly on US airlines between during the Thanksgiving holiday this year. Imagine each one of these individuals and billions of collective actions they took to step on that plane: from deciding on a date to fly and choosing an airline, to weighing the costs and determining lodging, there’s no other industry so ripe for big data-driven streamlining quite like travel. Artificial intelligence (AI) has begun to create a more personalized, efficient customer experience for many, but as anyone who paid $800 for their holiday flight and suffered through multiple delays before reaching their destination knows, there is a ways to go before AI automates travel down to simply “Point A to Point B”.
“A slower feedback loop — meaning travelers taking 2–3 flights per year vs. watching a movie every night — coupled with billions of potential flights, makes training the algorithm a much more complex task.”
—Isabella Patton, Hopper.
“Making personalized recommendations in the travel industry means dealing with data at a scale and complexity unmatched by most companies building recommendation algorithms, like Netflix and Amazon, with significantly less frequent feedback,” said Isabella Patton, Product Data Evangelist at Hopper. “A slower feedback loop — meaning travelers taking 2–3 flights per year vs. watching a movie every night — coupled with billions of potential flights, makes training the algorithm a much more complex task.”
Launched in 2015, Hopper is essentially like having a conversation with a digital travel agent that lives in your pocket. You opt in to “watch” a trip on Hopper, and the app sends you a notification when it’s the best time to buy. Hopper users watch 100,000 new trips each day on average, and last month alone the app sent more than 23 million push notifications.
Webscale companies like Google have used AI for years to analyze millions of travel options every day, predicting fares and tailoring results over time. Facebook is developing its in-house “DeepText” AI engine in an attempt to compete in areas like travel search. So for companies like Hopper, they see the real opportunity to differentiate by building trusted relationships with users through personalized recommendations. The company recently launched Flex Watch, for flexible travelers who want to keep an eye on a whole continent, for example. (It also helps that Hopper has been named the most accurate flight predictor in independent tests.)
“It’s an incredible, challenging opportunity to be a part of the travel industry, helping pioneer the way machine learning will impact a $500B market,” she said. “The level of trust and guidance we’re bringing to the table helps us rise above the noise of huge webscale providers.”
To build Flex Watch, the Hopper product and engineering teams took into account geographic graphs to understand how users from different origins view the world. Someone opening the app in San Francisco, for instance, would have a different experience from someone opening it in Montreal. Beyond that, they built a variety of signals into Flex Watch to train it to get smarter over time: Did someone consistently dive deeper into Europe? Or beach destinations? Did they only look for dates around major holidays? All of these signals help Flex Watch learn about each individual user and make more personalized recommendations over time.
“At the end of the day, people really want to know when the best time to fly and buy is. What’s different is that contextual information is making these recommendations increasingly more personal and relevant,” said Patton. “The first time you open Flex Watch will look very different from the 10th, 20th, even 30th time.”
The Return of the Travel Agent…Sort Of
HelloGbye is an app where you simply type or speak your travel plans into any device, and you’ll get an answer in less than 30 seconds. “Our platform is fully digital, meaning that there are no people that answer your queries, it’s all powered by our proprietary technology,” said Greg Apple, HelloGbye’s Head of Marketing. “It’s a very scalable solution — fast and efficient.”
It’s a level of intelligence that could redefine the travel industry.
The goal for the application’s natural language processing as the ability for the user to speak to it “as if it were a travel agent sitting there understanding what you want,” said Warren Stableford, Head of Content and Product Strategy at HelloGbye. “We take a complex, spoken query and pull information from around the world from many disperse sources and provide an itinerary for you in seconds.”
It’s a level of intelligence that could redefine the travel industry. The team started with only five developers in 2011 and focused on developing a user-friendly experience that would make HelloGbye’s machine learning and natural language processing feel delightful. Further differentiating HelloGbye is its socialization feature, which lets you build an itinerary with up to 8 travel companions. No more trading links via chat messages — you can coordinate it all via HelloGbye and find prices in real-time.
Beyond group travelers, the group most arguably in need of travel help is business travelers. Business travel often requires the most human touch. Personal preferences, preferred airlines, scheduling concerns, budget limitations and corporate partnerships create layers of complexity. Imagine all of this data being analyzed by an algorithm that can help a business traveler or executive assistant plan itineraries, schedule meetings and book travel quickly, with very little manual research. An AI that knows the company’s travel budget and whether you prefer a window seat, that sort of tech can free everyone from a lot of tricky, costly, time-consuming administrative work.
But don’t count out humans just yet, says Patton. “As more and more technology gets automated, it’s important to remember that at the end of the day it’s humans who are interacting with our products”, she said. “We see machine learning as a way to democratize the travel planning process — learning about your preferences and the preferences of those like you in order to make the best recommendations. If we have a machine doing this, it’s inevitably going to be cheaper and more widely available for everyone. What won’t go away, however, is that need for human connection. When all else fails, will there be a human on the other end of the line to help me if the worst situation occurs?”
“Over time, and as more and more interactions take place, machines will be able to serve those solutions as well. But until they’ve seen every problem in every way possible, there will always be a need for humans to problem-solve with empathy. I think this is core to human beings: a need to be heard and understood, regardless of industry or technology.”
Change is the only constant, so individuals, institutions, and businesses must be Built to Adapt. At Pivotal, we believe change should be expected, embraced and incorporated continuously through development and innovation, because good software is never finished.
Window or A.I.sle? was originally published in Built to Adapt on Medium, where people are continuing the conversation by highlighting and responding to this story.
