The Squeeze: Margins, AI Search Costs, and the Agent Race

Eric Marketts

Last week we walked three floors of the same building out of Rio — the PSS layer, the agent interface forming above it, the IATA program tying it together. We covered the retailing momentum from that AGM. The contracts, the pilots, the volume growth. This week we go back to Rio for the other half of the story. The half nobody puts on a momentum slide. The outgoing head of the industry's own trade body stood up and said airline profits are getting cut in half this year .

Steph Nell

And that number is the lens for the whole episode. Hold it in your head. Net margin of two percent.

Eric Marketts

Welcome back to The V1 Airline Retailing Report. I'm Eric Marketts.

Steph Nell

I'm Steph Nell. Three narratives today, and they're a squeeze. Story one is the margin — what Walsh said in his farewell, and what it means for every distribution decision an airline makes this year. Story two is a cost the stack was never built to absorb — AI agents running search at machine speed, with a real number attached. And story three is who's moving while airlines are heads down. The intermediaries. They are not waiting for the margin to recover.

Eric Marketts

Three pressures. One window. Let's get into it .

Eric Marketts

Narrative one. June 6 to 8, the 82nd IATA AGM in Rio de Janeiro. Willie Walsh delivered his final address as Director General before leaving to run IndiGo. He used it to lay out the financial state of the industry, and the headline number is stark. Net profits forecast to fall from forty-five billion dollars in 2025 to twenty-three billion in 2026. Net margin compressing from four point two percent to two percent. Fuel up roughly seventy percent year over year, adding about one hundred billion dollars to the industry's collective bill .

Steph Nell

And the operational backdrop is worse than the fuel line suggests. The supply chain backlog now exceeds eighteen thousand aircraft. Average fleet age has hit a record fifteen point two years. Airlines are flying without more than five thousand aircraft they were counting on for fuel-efficiency gains. So they're paying more for fuel and burning it in older metal at the same time. That's the squeeze before a single distribution decision gets made.

Eric Marketts

So is there a bull case in any of this?

Steph Nell

There is, and it's real. Demand is holding. IATA projects passenger growth above two percent. And look at the demand signal in the market — airfares are up around twenty percent on many routes and consumers are still booking. Even airline CEOs have said publicly they're surprised by how strong demand is. The carriers that have a functioning retailing engine — dynamic offers, bundles, ancillary revenue that isn't just bags and seats — absorb margin compression far better than carriers still living on base-fare yield. If there was ever a year that proved the case for Modern Airline Retailing, this is it. The margin pressure is the argument .

Eric Marketts

Okay. And where does it break down?

Steph Nell

It breaks down at the word "buffer." A two percent net margin is not a buffer. It's a rounding error. One shock — a fuel spike past forecast, a demand softening, a geopolitical event — flips positive to negative. And here's the cruel part for our audience. The right strategic move is to invest in retailing transformation. But the CFO running a two percent margin is going to fight that spend this year, because every dollar is contested and transformation doesn't pay back next quarter. The airlines that most need to modernize have the least room to fund it. That's not a strategy problem. That's a treasury problem .

Eric Marketts

So what's the listener supposed to do with a number they can't control?

Steph Nell

Here's the critical take, and it's the part of Walsh's speech that wasn't about fuel. His real message was systemic accountability. OEMs delivering late, airports constrained, air traffic management outdated, governments raising taxes — every one of those players externalizes its cost onto the airline, and the airline has no pricing power to push back. Retailing transformation is partly airlines trying to rebuild commercial control in a system where everyone else sets the rules. So the question that actually matters is timing. Can Modern Airline Retailing deliver enough margin uplift — through ancillaries, bundles, dynamic offers — fast enough to matter before the next shock hits? Right now that timeline is uncomfortably tight. And the next two stories are about forces making it tighter .

Eric Marketts

Narrative two. Skift, June 1, ran a piece called "The High Cost of Infinite Search." It opens with one example that I have not been able to stop thinking about. A man named Kiruba Shankar wanted one airline ticket. One. He pointed Claude Code at the Etihad website and pulled eight hundred and eighty-one thousand, seventy-six fare options. Every date in a month-long range. Every stopover. Every combination across four routes. For one ticket .

Steph Nell

And his own description is the whole problem in a sentence. He said he was "just a user who outsourced a job to AI to find the best option." He wasn't attacking the airline. He wasn't scraping for resale. He was a normal traveler who handed the search to an agent. The agent did what agents do. It didn't get tired.

Eric Marketts

And every one of those searches costs the airline money. Walk through why, because this isn't loading a static page.

Steph Nell

It isn't. A travel search checks live availability, live rules, live pricing, sometimes a bundle build, before it returns an offer. That's compute, and on richer content it's real compute. The whole economics of distribution rests on the look-to-book ratio. Ten, twenty searches go into a normal trip, and the booking at the end pays for all the looking. That math only works because of a human limit. People stop looking. That limit is the load-bearing assumption under the entire system, and it has been quietly holding the economics up for fifty years .

Eric Marketts

So is there a bull case here, or is this just a bill?

Steph Nell

There's a bull case, and it's about who moves first. This is solvable. Search caps, smarter caching, server-side filtering to cut downstream calls, agent authentication, dynamic pricing that prices the query. The airlines and platforms that build intelligent search optimization now — cutting compute waste without killing conversion — have a structural cost advantage inside twenty-four months. And there's an upside hidden in the threat. AI search patterns reveal intent signals that traditional booking flows obscure. Handle the traffic well and you learn things about demand you couldn't see before .

Eric Marketts

And the bear case?

Steph Nell

The bear case is the timing, and it rhymes with story one. Solving compute economics takes investment — new infrastructure, real engineering — at the exact moment margins are at two percent and the CFO is cutting. Most airlines are not thinking about compute-to-order. They're thinking about whether the NDC integration is stable. And here's the trap. NDC makes them more exposed, not less. The whole point of NDC was richer, live, dynamic content. Richer content is heavier to serve. So the carriers furthest along on modern retailing are the ones whose search is most expensive to run at machine scale. The reward for modernizing is a bigger compute bill .

Eric Marketts

That's a brutal irony. So what's the take?

Steph Nell

The critical take is one word: consent. This is cost imposed on the airline without a commercial relationship and without permission. The airline didn't deploy AI. Someone else did, and the airline pays the compute bill for a search that may never convert. We touched the agent layer last week — agents stripping your offer down to a row in a comparison table. That was about your brand. This is different. This is about your costs. The distribution stack was architected around a human who stops looking, and that human is being replaced by an agent that doesn't. Nobody is pricing adversarial or uncompensated AI traffic accurately yet. So here's the watch item. The first airline to publish an explicit AI traffic policy — rate limits, API terms governing agent access, or pricing tiers for machine-generated search — that's the signal the industry has finally internalized the problem. Until then, everyone's building a faster road and nobody's updated the speed limit .

Eric Marketts

Narrative three. Same Skift cycle, June 8. An interview with Bryan Batista, the CEO of Skyscanner. And the framing is what makes it a distribution story, not a metasearch story. Batista said he's exploring agentic booking to ease the handoff to third-party advertisers, and he thinks personal agents are part of the future of flight metasearch. So while airlines are heads down on a two percent margin and an exploding compute bill, the intermediary with the largest global flight-search footprint is repositioning for the agent era .

Steph Nell

And the corporate context matters as much as the quote. Skyscanner is owned by Trip.com Group. It just expanded into B2B through an acquisition. It's pushing harder into the U.S. market. This is not a company in retreat. This is a company investing into the exact transition airlines can't fully fund right now.

Eric Marketts

Give me the bull case, because an incumbent experimenting in the open is usually a good thing.

Steph Nell

It is. Batista isn't a tourist here — eight years at Booking, time at Tesla, and he's being honest about the history. Metasearch already tried to process bookings on its own site once. Google tried it, Kayak tried it, Skyscanner tried it, and it didn't take. Skyscanner reverted to pure price comparison in the early 2020s. So he's not pretending this is new. He's asking whether AI agents change the answer. And if a metasearch player makes the agent handoff genuinely cleaner, conversion could improve for everyone plugged into it, airlines included. Open experimentation by a serious operator beats vaporware from a startup .

Eric Marketts

And the bear case?

Steph Nell

The bear case is who ends up owning the moment of intent. Whoever owns the agent handoff owns the demand interface. And read Skift's own take on this — they point out it's not hard to imagine an LLM extracting Skyscanner's flight footprint straight out of Trip.com to accelerate building a travel vertical. So the intermediary layer here isn't fragmenting into a healthy market. It's consolidating, inside one of the largest travel groups in the world. Airlines spent a decade fighting to own the offer. That was last week's whole episode. This week's lesson is that owning the offer doesn't help you if a metasearch-turned-agent owns the instant the traveler decides .

Eric Marketts

So what's the action item for someone listening inside an airline?

Steph Nell

The critical take ties the squeeze together. Story one is margin. Story two is cost. Story three is the demand interface getting recontested while airlines have the least capital to contest it. Batista is moving. Trip.com is moving. The agent front end is being built right now, and it's being built by intermediaries, not carriers. So the question for an airline commercial team is uncomfortable and direct. Who in your organization owns the agent relationship strategy? Not the NDC roadmap. Not the loyalty program. The strategy for how AI agents find you, represent you, and hand a traveler to you. At most carriers that role does not exist. The intermediaries have already filled it .

Eric Marketts

Three narratives. One squeeze. Let me put it together.

Steph Nell

And the cruel timing is the whole point. Airlines have the least room to invest in distribution defense exactly when the distribution game is being recontested. The two percent margin isn't only a finance number. It's the reason the cost story and the Skyscanner story are dangerous instead of merely interesting. You fight for commercial control with the resources you have. Right now, airlines don't have many, and everyone else in the value chain knows it .

Steph Nell

If this episode was useful, send it to someone making distribution or retailing decisions in the next twelve months — these are 2026 decisions, not 2028 decisions. And if you're working on the hard version of this — an AI traffic policy, the compute economics of agent search, or an actual agent-relationship strategy inside a carrier — we want to hear how you're approaching it. That's the conversation this industry isn't having loudly enough yet.

Eric Marketts

Please see the episode details for links to the stories we covered today. We'll be back next Monday. I'm Eric Marketts. And if you find this content interesting, please subscribe to this podcast. It helps bring more content like this to you every week. If you have any feedback to share with us or just want to say Hi, drop us an e-mail at INFO at V-1 Advisory dot C-O.

Steph Nell

I'm Steph Nell. Thanks for listening.

Eric Marketts

Stay sharp out there .