How does Delta's AI pricing work?

My assessment of the vendor's paper & the market realities

Airline Revenue Management has in many respects always been what we today call a GPT (Generative Pre-trained Transformer), a type of Artificial Intelligence.

The “generative” bit was the demand forecast that RM created, estimating bookings and cancellations to come for all an airline’s flights. The “pre-trained” piece was the system’s evaluation of large sets of airline Revenue Accounting data, analyst influences and other sources. The transformer was the process of turning the data into a demand forecast.

That does not mean airline RM was exactly the same as ChatGPT, Perplexity or your own favourite Large Language Model. Those tools are much broader in scope, more powerful and consumer friendly than the airline tech. But the basic idea was not so different.

So when I heard that Delta is now pricing it’s tickets with AI my first thought was “nah, they’ve been doing that for yonks, mate”.

Of course it only took a few seconds to recognise that this was not the point. I then wanted to get to the bottom of exactly what they were doing, how it works and why it may be important.

That is what this article is about.

I have listed and attempted to answer five questions that get to the nub of exactly what is happening with AI pricing at Delta and possibly some other airlines too.

I have looked at what exactly Delta said and the standard set of benefits and objections all over the industry comment pages.

Fetcherr, the vendor behind Delta’s AI pricing platform, has published a helpful white paper* which I have studied. This and a few other points can help us figure out what is really going on behind the scenes.

* “Alien Intelligence in the Boardroom: Generative AI’s innovative new market strategies” by Uri Yerushalmi, the chief boffin behind Fetcherr’s whizzy algorithms

The white paper was published in December 2024. In the fast-moving world of AI this is aeons ago, so things will have moved on a long way since at the vendor. But since airlines move slow I am confident it reflects what is happening at Delta right now.

The five questions I have tried to answer

Q1. What type of AI is Delta using?

Q2. To what extent is the AI pricing continuous or discrete?

Q3. Are Delta’s AI pricing algorithms influenced by airline legacy or state-of-the-art pricing infrastructure in other high volume commodity markets?

Q4. What exactly is Delta’s AI estimating and what parts of RM is it leaving untouched?

Q5. What challenges will Delta encounter as it tries to make it’s AI pricing a success?