GenAI, the inflection point
The topic of artificial intelligence (AI) has been on the table for several decades. In recent years, this transformative journey has reached memorable milestones such as IBM’s Deep Blue beating a chess world champion, and more recently, DeepMind’s AlphaGo made headlines by winning against a professional Go player. Late last year, we’ve come to an inflection point, where AI got closer to conquering the Turing test, which discerns human intelligence from artificial one.
We are now living in a new era, brought by the revolution of Generative AI (GenAI), a type of applications that creates text, sound, or images (artefacts) from manual inputs (prompts). While AI has made significant strides across various domains, GenAI impresses as it mirrors human intelligence, particularly in generating articles or drawing images.
Simultaneously, the aviation industry has demonstrated its appetite for technological adoption, with a rich history spanning from pre-internet network connectivity with airports and travel agencies to cutting-edge advancements in automated aircraft piloting and other process automation like sorting billions of bags.
Will airlines and the air travel industry embrace GenAI? What impact will it have on customer experience and airline operations? Can passengers expect a GenAI revolution?
Humans and machines
The author of this article is a human being, who enlisted the assistance of ChatGPT-4 to compile this piece in an attempt to combine, rather than oppose, human and artificial, or humans and machines. I’ve used GenAI to guide the article’s structure, exploration, and referencing, producing the AI-enhanced content that you are reading now. For clarity, “AI” in this article refers to Artificial Intelligence, and not to the IATA 2-letter code for Air India.
My interest in artificial intelligence started in the 1990s when studying and programming neural networks using Yann LeCun’s method (Optimal Brain Damage, Le Cun, Denker, Solla 1989). LeCun noted, more than thirty years ago, that “as the number of parameters in the systems increases, overfitting problems may arise, with devastating effects on the generalization performance”. Today’s Large Language Models (ChatGPT being a LLM) may contain hundreds of billions of parameters to be trained on datasets containing billions of words, which raises the question of generalization. Are these LLMs good at memorizing or can they generalize and understand?
My interest grew when exploring AI applications in aviation in the 2010s, see for example IATA’s 2018 white paper about AI in Aviation (not available on IATA’s website any longer). Most of the findings are still relevant, from AI-enhanced customer touchpoint capabilities to operational capabilities and supporting capabilities.
Vision for the AI-powered airline
The vision for an AI-powered (including GenAI) airline is a travel service provider with enhanced safety, sustainability, customer-centricity, and operational efficiency.
Airline customers will benefit from all these enhancements in many ways, including smooth flights, on-time journeys, reduced or eliminated queues, personal travel advice and services, and more.
Value is in scaling human capabilities
AI offers airlines value by augmenting or replacing current human-dependent tasks. AI is scaling human capabilities.
GenAI, via the Language Models, has an obvious application for customer interaction, where AI could enhance interaction quality through email, web forms, and call centers. Early attempts at digitizing these interactions with chatbots fell short of human interaction capabilities. However, GenAI promises improved accuracy and response times, adding a human touch through cabin crew who deliver GenAI-driven responses.
GenAI has a broader ability to analyze large of text and data, and to produce analysis and recommendations. This capability has an application in commercial planning, which encompasses network and schedule planning, revenue management, and offer creation. As we’ve discussed in an article last month, AI can help shape personalized experiences at scale
Finally, AI can improve operational planning, from optimizing route planning to predictive maintenance and disruption management.
New paradigm, new solutions
When seeking solutions, airlines have three main options: license a proprietary solution, partner with a tech startup, or develop their own solution based on an open-source platform.
The pros of proprietary solutions are the integration with existing software and the perceived security. The cons are typically the speed of implementation, the innovation and the contractual limitations.
Tech startup would typically provide the opposite relationship, based on speed, innovation and flexibility, but lacking integration and scale.
Finally the development of an solution may seem to be limited to large airlines, with IT teams, but actually the barriers to entry have lowered thanks to the availability of cloud-based open-source solutions. The benefits of this approach include control, speed, agility but also tested security.
Hallucination and privacy risks
Although the benefits are substantial, potential risks associated with GenAI, such as hallucinations and data privacy, may be underestimated. “Hallucinations” refer to AI systems making statements that don’t align with reality, while data privacy issues arise when these systems learn from confidential user data and produce content in response that is made available to the public and competition.
For airlines, the safety risk is the potential for applications to make erroneous assumptions on routes, maintenance, or flight decisions. Given safety is the top priority in the airline industry, all AI applications that could affect safety will be under intense scrutiny.
Similarly, the risk for customer interaction arises when chatbots make incorrect recommendations, leading to customer disappointment. This can be mitigated by fact-checking GenAI outputs, while increasing staff productivity.
Lastly, there is a risk to airlines’ profitability if the GenAI systems are not properly managed or integrated. Many airlines prefer to test and learn, typically with internal use cases, such as editing documents and drafting emails. Some airlines have already identified all the GenAI use cases across their business, sorted by level of impact, effort and risk.
A revolution is coming
GenAI is clearly revolutionizing many aspect of the airline business and the customer experience. Companies that have been slow to prioritize technology or form a digital strategy risk falling behind the digital trailblazers that have embraced AI, and since the beginning of the year GenAI.
Airlines should be cautious of over-reliance on a single technology and/or partner for their GenAI experiments as it could lead to dependency, slow innovation or limited access to their own data.
Leveraging their extensive history with IT, airlines will undoubtedly make informed decisions as they navigate this new chapter in their technological journey. Customers can definitely expect a revolution in the travel experience in the coming months.
Article by Eric Leopold
