Delta Will Use AI To Cut Jobs And Set A Different Ticket Price For Every Passenger—CEO Says Profits Could Rise 50%

Delta CEO Ed Bastian says AI will increase their profits by 50% by cutting costs, removing humans from some loops, and raising fares. That’s the difference between 10% margins and 15% margins or about $3 billion.

The airline wants to replace slow decisions made by people with constant machine-made decisions across pricing, upgrades, crew recovery, maintenance, fuel and the back office. As United Airlines has already acknowledged while reducing management headcount, some people will disappear.

Bastian likes calling it “augmented intelligence” as a better marketing term than artificial intelligence, as he explained on Scott McCartney’s Airlines Confidential podcast:

When you think about what AI does, it allows you to hopefully make better decisions. And because you have more access to data, more timely, it gives you a clearer perspective of what the opportunities are. And our industry, our company, we have lots of decisions we have to make every day—whether it’s how we’re managing the daily operations or what’s happening in the maintenance realm and with our engines and the history of the engines.

There’s so much data around our business. Revenue management, you talk about, and pricing. So much data we don’t use but a small fraction of, because we don’t have the capability and we don’t have the programming. And so much of our technology is backward-looking in terms of historically what happened. We need forward-looking data, predictive data that has the intelligence to anticipate before issues become issues and try to make decisions better, and AI is going to provide that.

Now, it’s not going to be easy, and there’s no off-the-shelf solution to how you better manage crews and weather and fuel burn and engine performance. But all those things go into play in terms of running a great operation, both financially as well as from a customer standpoint.

Then he explained the payoff:

If we could take two, three, four points of our cost down from making smarter, better decisions over a series of several years—my gosh—you look at a margin at Delta, you go from a 10% margin to a 15% margin, it’s a 50% improvement in your profitability. These things are billions of dollars substantial. Or on the revenue side, because you can make better revenue decisions on how you manage different buckets. So I think the opportunity is there. I think the issues of trust and governance are also really important, and that’s why I never refer to AI as artificial. I always call it augmented intelligence. It’s going to make our people smarter and better.

Delta Explained To Investors What AI Pricing Looks Like

At the airline’s November 2024 Investor Day, then-President Glen Hauenstein described a complete reengineering of pricing. Airlines traditionally have one group set a grid of fares and another group decide which fare buckets remain available. Delta’s goal is to merge the two into “offer management”: one price generated for a particular shopping request at a particular moment.

Hauenstein described AI as a “super analyst” working continuously. Delta began by letting Fetcherr’s system control a small share of domestic inventory in a constrained environment, then expanded its testing. The strategic significance was never whether the first deployment covered 1%, 3% or 20% of fares. It was that the machine could keep learning and repricing while human analysts sleep.

Delta Gave Congress A Technically Careful But Misleading Answer

When lawmakers objected to “surveillance pricing,” Delta told Congress that it was not setting individualized fares using a customer’s personal data. Delta has continued to say that its current AI system uses aggregated market information, not an individual’s circumstances or prior purchases.

That’s true today but is contradicted by what they told investors: an offer available at that moment to “you, the individual.” The airline’s response collapsed two different questions into one:

• Is Delta currently feeding a named customer’s personal data into Fetcherr to calculate that customer’s airfare?
• Does Delta intend to move from static fare grids toward individually generated offers?

Delta answered the first and pretended it answered the second. The company was very specific about what it was not doing at that moment while its investor presentation was specific about where offer management was headed.

And an airline doesn’t need to know that your father passed away to infer willingness to pay. It can use route, date, time, device, sales channel, loyalty status, shopping history available in the session, whether you arrived through a corporate portal, and the behavior of statistically similar shoppers. Regulators and airlines can spend years arguing over which of those inputs makes a price “personalized.” The economic function is the same: estimate the highest price that wins this sale without losing the customer.

Laws Against Personalized Pricing Don’t Apply To Airlines

The FTC is warning retailers not to use private consumer data to raise prices, but the Federal Trade Commission Act expressly excludes “air carriers and foreign air carriers” from the Commission’s Section 5 authority over unfair or deceptive practices. That authority rests with the Department of Transportation.

And when DOT specifically considered this in 2014, in approving IATA New Distribution Capability efforts, they did not consider personalized airline offers to be illegal, leaving open whether differentiation based on income, marital status or trip purpose would be unfair or unlawfully discriminatory.

And while several states are passing laws on the matter, those are expressly preempted by the Airline Deregulation Act.

Personalized Pricing Can Produce Lower Fares

The reflexive political objection is that AI will identify a funeral traveler or stranded passenger and charge more. In that scenario, though, high willingness to pay doesn’t translate to high fares because if one airline wants to charge more, another airline sees the profitability of that customer on that trip and undercuts the first airline on price. And there’s a race downward back to normal pricing because there’s money to be made.

Just as importantly, personalized pricing can also mean lower prices. Airlines have always wanted to do two conflicting things:

• Collect a high fare from the business traveler who must be in Chicago Tuesday morning.
• Offer a low fare to the leisure traveler who will otherwise stay home, so an empty seat produces some revenue.

The problem with publishing the low fare to everyone is that the business traveler can buy it too. Traditional fare rules like advance purchase requirements, Saturday night stays and inventory buckets are crude fences meant to separate those customers. AI can build a more precise fence.

If a model becomes confident that I will not buy at $500 but might buy at $275, it can be rational for the airline to show me $275. That is incremental revenue, not a giveaway. Another traveler on the same flight may be shown $500. The airline earns more overall because it discounts where a discount changes behavior and avoids discounting where it does not.

That lets airlines offer more discount fares to more people, because they can do so without offering them to people who will pay more.

There Will Still Be Jobs

United’s chief financial officer said the airline’s management headcount had already fallen 4% and that it planned another 4% reduction as process changes and AI made headquarters work more efficient. That was unusually candid. The airline functions most exposed are not difficult to identify:

• Reservations and customer care: routine changes, refunds, policy explanations, social media responses and disruption rebooking.
• Revenue management: fare and upgrade monitoring, inventory changes, forecast updates and exception reporting.
• Back-office finance: accounts payable, revenue accounting, document extraction, reconciliation and fraud review.
• Marketing and loyalty: campaign creation, customer segmentation, offer selection and testing.
• Network and operations: schedule optimization, aircraft swaps, gate assignments and disruption recovery.
• Crew planning: pairings, rostering and re-pairing thousands of employees when weather breaks the schedule.
• Maintenance: predictive work planning, records review and technical-publication searches—reducing analyst work before it replaces licensed mechanics.

Pilots, flight attendants and mechanics are protected for longer by safety rules, licensing, unions and the physical nature of their work (although robots are coming, and that’ll affect ramp and baggage).

A crew controller presented by AI with three good recovery plans can make a better decision than one working through hundreds of constraints manually. A mechanic who gets the right technical answer immediately can return an aircraft to service faster. One single top employee can do far more, far more effectively with AI.

It’s a specific case of the broader phenomenon that AI will accelerate the productivity of the most productive people. It will also improve prospects for those at the lowest rungs. You don’t need to know cultural codes, to be able to write proper emails that are well-received, when the AI can do that for you. But those in the middle get squeezed. The easy ticket to the upper middle class that comes from becoming an average consultant, lawyer or accountant goes away. Just entering the machine is no longer winning.

That doesn’t mean fewer net jobs. Compute still costs money. We’ve known about comparative advantaage since David Ricardo, that as long as there are tradeoffs it doesn’t matter if AI is better at everything than people are still things that are more economic for people to do. As long as there are limits on energy and chips there will be jobs, but the jobs will be different.

Bastian’s Margin Math Seems Wrong

The argument here is that AI will improve price discrimination (and therefore total revenue), reduce cost, and improve resiliency. That will be huge for the business. But overall some of those gains will be captured by the passenger, rather than accruing entirely to the airline. Other airlines will be doing exactly what Delta is doing – United shed 4% of management staff last year, and expects to do the same this year – and with a lower cost base that results will compete down price.