We recently connected an AI agent to an airline’s live inventory and completed a real booking.

Not a mock.

Not a sandbox.

A real transaction.

Under the hood, this was powered by an MCP-style architecture that allows AI agents to interact directly with airline systems.

If you’ve been following the hype around AI in travel, that probably sounds like the milestone.

It’s not.

The industry is focusing on the wrong problem

Everyone is building:

• AI travel assistants

• Chat interfaces

• Trip planners

But very few are solving the thing that actually matters:

Can the AI complete a real booking on real airline infrastructure?

Because that’s where things break.

Airlines don’t have a UX problem. They have a systems problem.

Turning “book me a flight” into a valid ticket requires:

• Real-time availability

• Pricing logic

• Fare rules

• Passenger data validation

• Payment + ticketing

That entire chain has to work perfectly.

And it has to work inside systems that weren’t designed for AI.

This is why most AI travel demos stop before checkout.

This isn’t new. It’s the same problem airlines have always had.

Airlines don’t make much money flying passengers.

In fact, the industry is expected to make around $7.90 per passenger on average .

That’s not because demand isn’t there.

It’s because the infrastructure behind the transaction is incredibly complex and fragmented.

So airlines compensate:

• Loyalty programs

• Credit cards

• Ancillary revenue

Not because they want to.

Because they have to.

AI doesn’t fix this. It exposes it.

AI makes it easier to generate intent:

“Find me a flight”

“Book the cheapest option”

“Change my return to Friday”

But intent is useless if it can’t be executed.

So the real shift isn’t AI interfaces.

It’s what sits behind them.

The real opportunity: the transaction layer

If bookings move into AI assistants, the question becomes:

Who actually completes the transaction?

Not:

• Who owns the interface

• Who generates the demand

But:

Who can reliably turn AI intent into a valid airline booking?

That’s a different layer of the stack.

And it doesn’t exist yet in a clean, scalable way.

Why this matters for airlines and PSSs

Over the next 12–24 months:

• AI assistants will become a real distribution channel

• Airlines will be asked: “Can your system support this?”

• PSS providers will need to prove compatibility

This won’t be optional.

It will be table stakes.

What we’re building toward

The demo isn’t the product.

It’s proof that:

• AI can connect directly to airline systems

• Bookings can happen without an OTA

• The airline can still own the transaction

But more importantly:

It highlights where the industry needs to go next.

Because booking is just step one.

Servicing, changes, disruptions — that’s the next frontier.

AI booking is coming.

But not because of better chat interfaces.

Because the infrastructure is finally starting to catch up.

And the teams that understand that layer—

not just the UI—

are the ones that will shape what comes next.

Want to see how it works? → Book a demo

If you’re a traveler:

→ Join the waitlist and get early access.

PS: Want to see where this is going? Sign up below and get an early look at how we’re rebuilding airfare booking from the ground up.