Travel executives can create an actionable AI roadmap for 2023 by understanding how the technology works, its varying practical applications, and how it will affect their organizations on a broader scale.
This sponsored content was created in collaboration with a Skift partner.
It’s the “Year of AI,” and artificial intelligence (AI) is everywhere — including in a growing number of products and solutions for the travel industry. But AI isn’t a monolith. For all the promises attributed to the concept, it’s not always easy for executives and other decision-makers to understand what they should be deploying, how it affects their current systems and processes, and how it will ultimately benefit the bottom line.
By exploring how AI works, its varying practical applications, and how it can exponentially scale data intake and analysis, travel companies can better understand where they should take their AI roadmaps in 2023. Here are three things every travel executive should be thinking about as they embark upon that journey.
Focus on Practical Applications
The news has been abuzz about ChatGPT, DALL-E, and other so-called generative AI programs that can create new, unique outputs based on specific prompts they’re given. It’s an exciting space that has deep relevance for the travel industry, but generative uses are still in the early developmental stages. Today, it’s important for executives to understand that AI comes in many different forms.
“[Generative AI] is an interesting space, but we are not there yet for airlines,” said Kartik Yellepeddi, vice president of ML and AI strategy for FLYR Labs. “You can’t expect to generate new pricing strategies out of the blue… yet.”
In the travel industry, “supervised” uses of AI are much more controlled than the generative applications that have been popular in the news, Yellepeddi said.
So how does a supervised learning model work? Using airline revenue management as an example, an AI model will label historical outcomes for pricing as “good” or “bad” based on how given actions contributed to the ultimate goal of maximizing revenue. The AI can then assess new variables and suggest pricing modifications consistent with those “good” decisions. Through thousands of inputs and repetitions of this action daily, it trains to do more of the good and less of the bad and becomes smarter as time progresses. And at some point, it’s learning from itself.
“AI is data hungry, but the advantage is that it’s highly scalable,” said Yellepeddi. “The art is in how you design it, and it can theoretically learn any task you give it. If there is a pattern out there, it’s able to learn that pattern and recommend what action to take to maximize the ‘reward’ it’s trained to seek.”
To understand this better, take the case of optimizing the price for any given flight, which typically opens for sale 300 days before departure. Every day, thousands of variables, such as new bookings, changes in search volume, competitor sales, and pricing changes, affect the flight’s potential price and final outcome. AI can analyze this constantly shifting context in a way that’s impossible for a human to do independently, providing pricing analysts with a depth of information that was not previously available.
How to offer ancillary products, when to overbook, how to price cargo space, and how to deploy marketing dollars are other ways that airlines and travel companies can take advantage of AI models to improve their decision-making.
“We realized that airline pricing and forecasting was a generic use case and that you can apply the same machine learning technology to a number of other important commercial functions,” said Yellepeddi.
These types of practical, day-to-day uses allow companies to dip their toes in the water and deploy AI capabilities while operating in relatively low-risk scenarios.
Seek Solutions That Can Scale
As travel companies look to take advantage of AI opportunities on a long-term, organization-wide basis, they must be ready to invest time and technology into shifting how they operate.
For example, revenue management systems have been historically built on fixed-growth scenarios, looking broadly at year-over-year changes.
“Historic revenue management systems were tasked to do one thing — price the flight — and now there are 10 or 15 transactions occurring with the same customers during the same trip, from ancillaries to other offers,” Yellepeddi said.
The rate of change has significantly accelerated in today’s travel environment, and AI can become an asset by being far more dynamic and reactive than humans. Cloud technology allows companies to be more flexible in their data storage, analysis, and application, while legacy systems with fixed servers aren’t built to scale in this way. Because fixed servers have fixed costs and fixed capacity, that means companies are unable to allow the AI the freedom to use all the available data, because they have to make predetermined decisions about how much information they can reasonably manage. That, in effect, hinders their ability to scale and take full advantage of running sophisticated AI models.
“Cloud has really changed the landscape,” said Yellepeddi. “Most importantly, it allows you to use all the data in the decision-making process.”
Trust the Technology
One of the most important things for executives to consider when using AI is that they’ll have to relinquish some level of control and trust the technology.
If there are thousands of data points related to pricing generated every day, analysts might be reasonably able to look at a few hundred of them. It’s the responsibility of the AI not only to look at all of those data points but also to flag which ones require human attention to drive meaningful outcomes. Building a deeper level of trust enhances analysts’ abilities to use the information to optimize their recommendations.
The strength of AI is not necessarily the ability to be right all the time but rather its ability to react to situations quickly, continuously explore and exploit market opportunities, and learn from its mistakes faster on a larger scale than humans.
From that perspective, an important recent evolution of AI has been improved explainability — that is, being able to “show its work.” AI models aren’t just spitting out decisions, they’re now also able to provide information on how they came to those decisions.
“If the ‘good’ decisions outweigh the ‘bad’ ones in the end, how it makes the revenue should matter less — as long as you have visibility into the decision-making process when needed,” Yellepeddi said. “It’s important to build that trust to drive adoption of any advanced AI technology.”
AI’s effect on productivity, its ability to exponentially scale data analysis and decision-making, as well as learn on the job, show its work, and escalate to humans when required — will drive automation, efficiency, and profitability across the travel industry, from revenue management and marketing to cargo, maintenance, and more. By taking advantage of practical commercial opportunities today, executives will also set themselves up to understand and integrate cutting-edge AI applications as they come online in the near future.
For more information about FLYR and its commercial intelligence and optimization platform that leverages AI and deep learning, visit www.flyrlabs.com.
Have a confidential tip for Skift? Get in touch