AI & Airline Distribution, Two Scenarios
Game Theory suggests that the future of travel retailing is dismal
The future of airline distribution will undoubtedly be powered by AI in some way or another. Unfortunately that is no guarantee that what emerges will actually benefit either airlines or consumers.
There are two scenarios that I see. The first is airlines and their tech vendors finally fulfilling the promise of NDC and offer-order. They will create personalised travel marketplaces offering consumers what they really want, or something close to it (see this article and this article).
On the other hand, the future might be dismal. Airlines and travel retailers could end up putting their resources into nothing more than robots screaming at each other.
Unfortunately a spot of Game Theory*, a branch of maths examining strategic interaction, suggests that we get might get the latter in practice. Read on to find out why…
* Specifically the Prisoner’s Dilemma game: I will explain what this is later in the article
Challenge One: It will be hard for airlines to achieve revenue value from AI in the near term
Civil aviation has a great track record of innovation. Planes fly further, engines burn less fuel and passenger whizz across the globe in flying beds.
Prices are lower than ever too, in real terms, especially when loyalty comes into play.
My Club World (Business Class) flight from Hong Kong to London late last year cost only GBP 35.10 (USD 47.03), thanks to On Business points I did not even need to burn any of the Avios loyalty currency. It really was an almost-freebie, a tremendous achievement for airline economics.
Unfortunately, generative artificial intelligence is unlikely to unlock a sack full of revenue for airlines, at least in the near term. There are four reasons why (see article):
1. Sheer inertia ensures that even the whizziest of algorithms take a long time to deliver on their promises because they are built on top of legacy platforms
2. Cost, especially salaries as airlines have a track record of overpaying for technology and underpaying for staff (see article) so will not be able to hire the people they need to make things happen
3. Messy data, as the oodles of bytes in airline systems are not linked, plus much is at best misleading or at worst corrupt (see article)
4. Airline commercial systems themselves will at best struggle to cope with AI on top and at worst they might collapse under the weight, like London-based BA’s did in 2022 (see article)
Challenge Two: Predicting what consumers actually want to buy will be tougher than it sounds
Airline travel marketplaces are supposed to offer consumers a wide range of travel-related products and services with their flights. These might include everything from restaurants and meeting rooms to bungee jumping over the Grand Canyon.
Turkish Airlines holidays is probably one of the world’s most advanced airline travel marketplaces at the moment. But it still has a long way to go until it delivers truly personalised offers across a long tail of travel-related services.
For example, hot air ballooning in Cappadocia is currently featuring heavily in Turkish Airlines advertising. But I was not able to book a flight, hotel and hot air balloon package on their holidays site. One day, maybe even one day soon, I will be able to.
There is a simple test to check whether or not an AI-powered airline travel marketplace is delivering what consumers want:
1. It should recommend McDonald’s when people fancy a Big Mac
2. At the same time it should recommend Le Manoir aux Quat Saison, a fancy eatery in leafy Oxfordshire, when people going to that part of the world are in the mood for lobster and foie gras
Call it either the Manoir-McDonald’s test or the burger and lobster test.
Today airlines would fail the burger and lobster test without question. In some but not all cases, tech giant Google might not.
Put simply, not everyone has the same idea of a good time, what constitutes luxury or what delivers value. For some, a suite with a view of the Eiffel Tower might represent the romantic break of a lifetime. For others, it might all be about a heart-shaped bed.
This does not mean that airlines need to offer consumers the one choice which they believe is the most optimal. Well-stocked airline retailers should have attractive “window displays” and shelves full of compelling travel products to inspire people to shop there.
Successful airline retailers will need variety, social proof and flexibility.
McDonald’s may belong on a shelf with Five Guys and Byron Hamburgers, other suppliers of tasty patties.
On my recent holiday to San Francisco a travel retailer would have done well to recommend Annie’s hot dogs. Their jumbo beef dog is so good I ate four of them in ten days!
The trick will be ensuring that McDonald’s or Annies should not appear on the same shelf as Le Manoir.
Meanwhile on the same trip to San Francisco, it turned out though that Fisherman’s Wharf was not up my street. An absolute hotspot and must-see for many tourists in San Francisco, most travel marketplaces would offer it almost by default.
However once I had been to Fisherman’s Wharf I realised there were good reasons why it was not for me. A truly amazing airline travel marketplace would recognise this and either not bother with Fisherman’s Wharf or put it on a low “shelf” that is harder to see.
Airlines creating such an experience marketplace will need to figure out how to play around with the options, trading off flight price against other travel spend and any commission earned from partners (see article), a shopper’s previous brand loyalty and many other attributes. This is going to be tough.
The serious question of whether or not airlines really have the data they need to personalise offers also needs to be considered.
The technical challenges of scraping data from all over the Internet and matching it to specific consumers in an airline’s database will be formidable. It is hard to see this getting any easier if the public become more privacy sensitive or the regulatory frameworks like GDPR become stronger.
Challenge Three: Hotel retailers and AGIs are also on the case
Expedia, Trip.com and Booking.com are three of the largest hotel retailers in the world. They are all starting to incorporate AI-driven platforms into their offers, which will compete with airline marketplaces.
Expedia markets their Romie as an “AI-powered travel buddy”. Trip.com’s service is called TripGenie and Booking.com offers AI Trip Planner.
Artificial Generative Intelligence (AGI) firm OpenAI might also be able to compete with the incumbents. Their Operator AI is designed to search the web like a human.
It is hard to see the technologies that airlines have developed like offer-order and NDC competing with agile and nimble entrants like OpenAI.
Challenge Four: Vendors are ready with non-solutions
Parkinson’s Law of Board Meetings states that the time spent on discussing an item is directly proportional to how much the directors understand it. So the discussion about providing chocolate or plain biscuits in the pilot’s break room goes on for hours.
Parkinsons’s Law also states that the time spent is inversely proportional to how much a project costs. So Amadeus Nevio, PROS AI Agents and SABRE Mosaic, which the directors do not really understand and are really expensive, go through on the nod.
SABRE’s Mosaic advertising currently includes a lady and a gent holding surfboards. If you think about it, airlines should be perfectly positioned to sell surfboards. Travellers going on holiday to the beach may well be after a surfboard. The airline should sell them with the flights.
The reality is different. If any airline anywhere has ever sold a surfboard using SABRE’s technology I would be surprised.
Fittingly, SABRE is using the surfboard picture to talk about disruption management. A good topic for another day…
All the above means that the Prisoner’s Dilemma Game is bad news for airlines
Each year industry shibboleth ATPCO holds a shindig called Elevate. This year, I am told, a speaker said “we are no longer designing for people, we are designing for their proxies”.
This means that airlines are building robotic offer-order generators that create hundreds of different products to choose from. And it is not people doing the choosing, because people can only consider two or three or four products before they get bored. Instead it is robots behind travel marketplace software.
So what will this future look like. It could be a set of beautiful travel marketplaces that genuinely do offer consumers a wide range of choice. This will be expensive. And the four challenges I outlined above will make it difficult if not impossible too:
1. Using AI to generate revenue is not necessarily practical in the near term
2. Predicting sets of choices that resonate with consumers will be tougher than it sounds
3. Hotel retailers and firms that are faster and better at AI than airlines are on the case
4. Vendors promise much but the majors have a track record of delivering little.
So what of the alternative?
I am reminded of the Prisoner’s Dilemma in Game Theory. In something like it’s original form it goes like this:
1. Two criminals are caught by the police
2. There is weak-ish evidence that both are guilty
3. If neither criminal confesses, they both go down for two years
4. In a plea bargain system, if one criminal confesses the confessing criminal gets released and the other criminal goes down for ten years
5. If both criminals confess, they both go down for nine years.
There are four possible outcomes:
1. Neither criminal confesses
2. Both criminal confesses
3. Criminal 1 confesses and criminal 2 does not
4. Criminal 2 confesses and criminal 1 does not
The outcome where neither criminal confesses is not, in economics terms, an “equilibrium” because both criminals are going down for two years. But if they confess and the other does not, they get off. So they confess.
The two outcomes 3 and 4 where one criminal confesses and the other does not are also not stable. Given that the one criminal is confessing anyway, the other can reduce their sentence from ten to nine years. So they confess too.
The only stable outcome is that both criminals confess and both go down for nine years. They would have been better off (only two years each) if neither had confessed.
I think that what we are seeing with AI in airline distribution is a bit like the Prisoner’s Dilemma game. But with airlines and agents choosing to either “one-up” the other or maintain the status quo instead of criminals confessing.
Airlines and agents can both stay where we are today, which serves consumers adequately. Or they can invest heavily in AI to drown the other in a choice of offer-orders. And given that either airlines or agents are investing heavily in offer-orders, the other will decide to do that as well.
At the end of the day all we will have is a large number of expensive AI robots screaming at each other.
Airlines and agents will be worse off as they need to pay for this useless technology. Passengers will be no better off because somebody still needs to sit in seat 58J and the meals will still be limited by what can actually be prepared in an airline galley.
So what is the solution?
The Prisoner’s Dilemma can be solved more amicably if the game is repeated. That way, the criminals learn to collaborate and neither party confesses. That is bad news if the “criminals” are dodgy geezers. It is, however, good news is the “criminals” are states armed with nuclear weapons.
I think the challenges of airline distributions need a solution too. It is not going to come from more and more technology. Instead it needs to come from airlines embracing the human element. They should price for humans, not machines. The first thing to do is to shift priorities from investing in technology to investing in people.
Changing the topic!
What should Global Airlines do with their A380?
Last week’s Global Airlines flight from Glasgow to JFK was at one time the most followed air mission on FlightRadar24, a plane tracking site.
I have not commented on Global Airlines before because I am not entirely sure what they are trying to achieve.
The company may be an airline in the branding sense. But it may not really be an airline in the technical sense. It lacks an Air Operator Certificate, a Type A Operating Licence and the ability to sell it’s own tickets.
In effect (as I understand it) they bought the A380, dry leased it to Portuguese/Maltese carrier Hi Fly, whose crew operated the flight, and wet leased it back.
Nonetheless they have managed to ship an A380 (see article) across the Atlantic and back in their own house colours with paying passengers on board. And that is more of an achievement than most of us in aviation ever manage. For that I give them a hearty cheer, a slap on a back and a well done.
But what should they do with their A380? The North Atlantic probably does need more competition, but I doubt the A380 is the plane to do it, especially on those relatively short flights to the east coast where market lore says that high yielding travellers prize frequency.
What I would like to know is whether Global could somehow use the same arrangement with Hi-Fly to get their A380 down to South Africa and back on a regular basis.
There are four reasons why I like the idea of South Africa Johannesburg and Cape Town as a destination for Global:
1. Currently under-served with high fares – good luck getting a Business Class ticket on a direct flight from London to Cape Town for less than GBP 3.5k, even in the sale – Global might not be able to get what BA gets, but with a decent set of group and bulk deals they would sell a lot of seats at decent yield
2. Johannesburg is an existing A380 maintenance base with low cost, helping Global put things right when they go wrong in an affordable way
3. According to my estimates based on fares data, BA probably earn at least $800k per flight with their A380 to Johannesburg (see article)
4. The market rewards a long period of aircraft down-time to achieve overnight flights in both directions so passengers can sleep rather than “wasting time” on a plane.
Can it be done? Answers on a postcard please.
Aviation in Eurovision
There were two awesome aviation references in the Eurovision Song Contest (see article) last Saturday.
Check out the Estonian entry to see Tommy Cash’s dancing plane. Once seen, never forgotten!
And take a look at Spain’s performer Melody doing the “el helicoptero” right at the end of her act. Viva la diva!!
Read more on Airline Revenue Economics: the Bob Branston saga of an imaginary airline CEO using AI to set up Vestal Atlantic
AI & An Airline Startup – Part One
AI & An Airline Startup – Part Two