Airlines are ramping up their exploration of generative artificial intelligence, and harnessing the technology in more creative ways.
Most of the focus is on Large Language Models—machine learning that can comprehend and generate human language text. Booking assistants, for example, are in fashion as airlines transform into online travel agencies.
Subscribe to our newsletter below
Recent launches include KLM’s Ask Atlas, an AI-powered recommendation platform that lets users design their perfect vacation destinations by selecting interests. Alaska Airlines also operates an AI tool enabling users to ask for suggested flight destinations based on a general topic of interest.
Meanwhile, Delta Air Lines unveiled generative AI-powered assistant Delta Concierge at CES in Las Vegas, and Lufthansa this month integrated its AI-powered booking assistant Swifty into Japan’s LINE messaging app so customers can book flights directly.
Air India also rolled out eZ Booking at the end of last month. This system allows users to complete their reservation on its website in fewer steps than currently available, by texting or talking to an AI Agent about their planned itinerary.
Trend watching
The goal of these LLM-based innovations is to boost profits. Air India’s eZ Booking, for example, builds on its AI.g chatbot. According to Microsoft, which helped build it, AI.g now handles 97% of customer queries, and it saves the airline millions of dollars in support costs. The chatbot has also answered seven million queries since May 2023, the airline said.
But airlines equally look to generative AI to make better decisions, from route planning and baggage handling to repairs and maintenance. In fact, any way to become more efficient.
But it’s also valuable as a means to gain an edge on competitors. Lufthansa has taken an innovative approach to stay ahead of the curve by teaming up with Germany’s Anacode. Together they built the “Airtravel Trend Insights Platform” which is a system that leverages LLMs to analyze trends, competitor activity, and customer preferences, generating strategic innovation ideas and recommendations.
For example it can detect current consumer trends, such as veganism. “Excitement is high, but focus is key—not every topic can be addressed. Each trend has a detailed profile to decide whether it should get a ‘seat’ in the airline’s strategy,” Anacode said.
Going beyond ‘mimicry’
Israeli startup Fetcherr sees potential in numbers, not words. It uses Large Market Models, or LMMs, which analyze and predict complex market dynamics in real time.
Unlike traditional language models, LMM leverages non-natural data like financial and business environments. Experts liken it to OpenAI's DALL-E, which creates images based on descriptions. In the same way, an LMM optimizes decision-making processes.
Through this approach Fetcherr, a PhocusWire Hot 25 Travel Startup for 2023, provides an AI-powered pricing and inventory control engine to airlines. Its co-founder believes there’s an element of hype around this latest wave of innovations, such as booking assistants, which “mimic” humans.
“Most of the buzz is around language models. That drives some conception of an agentic AI or of pieces of software that talk to you or help you, by language, do things,” said Uri Yerushalmi, co-founder and chief AI officer.
“We will see tons of those, because of the hype … but our vision is a little different because our basic core model is not a language model, it's a market model. Our conception is less around mimicking humans or talking to humans, it's more around restructuring the whole concept of decision making.”
Before launching, the company said it had spent three years training its engine with millions of data points from the airline industry. Now, bolstered by $90 million in Series B funding, the company is putting that into practice, and was name-checked at Delta Air Lines’ investor day event in November last year.
“This is something I'm really, really excited about,” noted Delta president Glen Hauenstein, referring to generative AI. He disclosed 1% of the carrier’s network was now being priced by the Fetcherr.
After he spent time working in trading and capital markets, Yerushalmi said when the company was started, it could have applied its technology to any industry. But it opted for aviation as it was ripe for disruption after decades of neglect.
“We found the airline industry quite complex because of all of the joint ventures and interlines and connections,” he said. “This complexity made the industry undisrupted for decades. Because of the complexities, a lot of solutions now in the market are based on equations or systems from the sixties or seventies. It made the market ‘thirsty’ for new technologies.”
Bad press
Some reports suggest AI-driven pricing can lead to higher fares, but the co-founder plays down the negative perception.
“On average the prices are slightly lower, because current systems are not very good at exploiting all of the capacity,” he said. “Usually the load factors are not high enough. AI is very good at filling the last seat, with a relatively good price. That's why on average the prices are slightly lower, meaning the passengers on average are paying less, but the revenue of the airline is higher.”
Maximizing loyalty
Another area AI has potential to improve is loyalty—which according to industry expert Cory Garner is the final of “three eras of distribution change spread across the last three decades”.
“Now the industry is in a third era which started after COVID. At this point airlines are looking to prioritize their loyalty programs, including credit card reward programs,” he said during a session at The Beat Live event in New York in December last year.
Airline pricing intelligence platform Airnguru argues airlines should leverage AI to unlock insights around loyalty programs and apply the rigour of revenue management to their reward schemes—or risk leaving money on the table.
“Airlines sit on a wealth of information within their loyalty databases. By aligning this with consistent pricing strategies, they could steer rewards to low load factor flights to minimize displacement costs and incentivise reward demand in underperforming flight clusters,” said Sergio Mendoza, co-founder and CEO. “This would allow them to incentivise travellers to book additional trips, instead of redeeming their reward tickets for the flights they had already planned on booking.”
Airnguru is now looking at “conceptualizing” this among its airlines customers.
Question of timing
Such is the speed of product launches, there’s a risk native-generative AI tech players will beat others to the game. For example, Air India’s eZ Booking was announced January 22. A day later OpenAI released “Operator”, an artificial intelligence-powered agent tool that can interact with websites by typing, clicking and scrolling.
Yerushalmi also said Fetcherr is gearing up to expand, offering its services to other travel industry sectors.
“We are now exploring other verticals,” he said. “Hotels is one of the verticals that is possible, which is similar to airlines in some senses. At the same time, it's much more fragmented.”
OpenAI, Fetcherr’s potentially agnostic solutions and China’s launch of DeepSeek point towards a new reality for airlines: disruption will always be just around the corner.
When asked about the other ways Air India is exploring AI, and how it keeps up to speed, Satya Ramaswami, chief digital and technology officer, said: “We envision Air India to be an AI-infused company. We expect AI to be all pervasive in our operations in a way that enhances customer experience and reduces our costs. We have deployed AI in various enterprise use cases and will continue to enhance our digital customer experience with AI capabilities.”
He added the carrier also has a presence in Silicon Valley and connects with the academic institutions and startups there to keep the airline up to date on what’s coming.