Aviation: Is AI Ushering in a New Era of Pricing?

It’s with this rather alarming headline that Hugo Pellegrin from Tom Travel signs his article. He relays the news that Delta Airlines plans to expand the deployment of its new AI algorithm, which combines a « Large Market Model (LLM) » , a « Generative Pricing Engine (GPE) » and a « Generative Inventory Engine (GIE) ».

Two U.S. Democratic senators have raised concerns, calling the initiative an « abusive practice » : « Individualized pricing, or surveillance-based pricing, replaces a fixed or static price with prices tailored to each consumer’s willingness to pay. Delta’s current and planned practices of individualized pricing raise not only data privacy concerns but are also likely to lead to price increases (…) ».

The topic is also worrying because the details of this individualized pricing have not been disclosed, particularly regarding the degree and nature of personalization. So, what should we make of this?

The issue is complex, and the challenges are of various kinds:

• The Legal Aspect

The key question here is about discrimination. Differentiating prices based on past purchasing behavior or web browsing – does that fall under segmentation or discrimination? Until now, demand has been analyzed by homogeneous customer groups: Early bookers vs. Late bookers, Leisure vs. Business, Groups vs. Individuals, One-Way vs. Round-Trip, Point-to-Point vs. Connecting flights, Domestic market vs. “End-of-line” markets, etc.

Price differentiation among these groups, without distinctions of nationality or gender of course, is considered valid segmentation. The legal question that will arise is whether Pierre and Aurélie, for the same round-trip, at the same time, on the same flights, can have different prices.

For example, Pierre might book on his PC after much hesitation and multiple website visits, first clicking on the Promo tab. Meanwhile, Aurélie books on her tablet, entering her dates right away without apparent flexibility, as if she had a constraint making her trip non-negotiable. Pierre might be considered, rightly or wrongly, more price-sensitive than Aurélie. But is that a legitimate argument for differentiating their prices? Lawmakers will have to decide. But at first glance, this looks a lot like discrimination.

• The Marketing Aspect 

Companies will need to justify offering a higher price to one customer over another, while looking them straight in the eye and saying: « We believe you are less price-sensitive based on your browsing history and past purchases. So, for you, it’s €1,200. Thank you.»

Or they’ll have to lie to the customer. Good luck to the Marketing teams, especially the Customer Relations teams.

Beyond the legal aspect, ethics matter in our profession. A simple, common-sense rule should apply to all: be able to explain to a customer, face to face, how the price was determined without trembling in your voice. If you can’t, you need to reconsider your approach.

• RM & Pricing Efficiency 

The effectiveness seems proven on Delta’s side since they plan to expand their algorithm’s deployment to 20% of their scope, up from 3%. Nevertheless, this raises several questions:

• • It’s assumed that the algorithm will combine individualized pricing with a traditional Yield approach. This means it will mostly apply to low and mid-season flights. On peak flights expected to be full, everyone is supposed to pay high fares, and there’s no room for a hesitant Pierre to be offered a low price – he would take the seat of an Aurélie who could pay more. So, the algorithm would be more relevant for flights that are underfilled, aiming to charge high prices to less price-sensitive customers even if Yield management is broadly opening up low fares. However, this is already widely practiced through current segmentation, where tools like Fare Families or Branded Fares allow well-segmented customers to pay high prices even on lightly booked flights, for example, because they book a round trip during the week (Sunday Rule) or seek flexibility. What real added value will this extreme differentiated pricing bring compared to the current, already highly sophisticated segmentation in aviation?
  • What about Distribution? It’s clear that on Delta’s website, Pierre and Aurélie won’t see the same price, but in GDS systems? If Aurélie realizes that agency prices cannot be differentiated, or at least not in the same way, she might choose to go through a distributor to avoid being overcharged. What about managing price parity? How to avoid cannibalization? The impact on Distribution is arguably the major issue of individualized pricing.

It’s likely that, knowing such algorithms exist, some clever folks will find loopholes and exploit them: creating fake profiles and purchase histories, publishing browsing guides guaranteeing eligibility for lower fares, etc. Startups have always emerged to exploit the gaps and excesses of RM & Pricing strategies. This will happen here too, inevitably. An Aurélie hiding behind a Pierre.

Beyond the legal aspects, this practice, whose efficiency seems questionable or at least fragile, relies on a shaky foundation: the belief that customers are absolutely unique and therefore deserve a different price each. A price per head. Or rather, a price based on the customer’s face. Marketing teams will tear their hair out. And algorithms will take control.

There’s no human left. Only customers and machines. It doesn’t exactly make us dream…