A Pacific carrier I worked with had 12 systems that mattered to the commercial team. The passenger service system. Revenue management. The GDS connections. Fare filing. Loyalty. Revenue accounting. The email platform. The data warehouse. None of them had been bought to work together. Each had been wired directly to the 4 or 5 others it touched, one integration at a time, across 15 years.
Nobody chose that. It accumulated.
Then a distribution mandate landed—NDC—and every one of those point-to-point connections became a place something could break. A fare-brand change in one system rippled through 4 others. Work that should have taken a quarter took a year, because the team spent most of it untangling integrations instead of shipping the capability.
The vendor answer was a platform. One system to replace the mess and future-proof distribution for good. It was a serious product. It would also have been the 13th system wired to the other 12—and the mandate after NDC would have found it just as brittle.
We did something less impressive and more durable. We built two things the carrier owned outright. An analytical data layer that every system fed once. A common services layer that everything else called, instead of calling each other. Integrate to the layer, not to the neighbor.
That sounds like plumbing. The result wasn't. The next capability the team wanted to ship—one that under the old wiring would have rippled through 5 systems—now ran through the layer. Time-to-market went from a quarter to a few weeks. They replaced their marketing platform 18 months later, and the migration touched 1 connection instead of 5.
Here is the part most roadmaps get backwards.
Future-proofing is a bet that you can see the future. You can't. NDC was going to be the future of distribution. Now agentic AI is rearranging the question before NDC has finished arriving, and MCP—the protocol that lets an AI agent query a system directly—is quietly becoming the way machines will ask airlines for fares. Every architecture hard-wired to last year's certainty is being rebuilt again this year.
The carriers that adapt are not the ones who guessed the standard right. They are the ones who own the seam the next thing plugs into. When you own the layer, a new standard is a new connection. When you rent the platform, a new standard is another 18 months and another check.
Composable beats monolithic for the same reason. A layer you own gets reshaped one piece at a time. A platform you rent moves when the vendor's roadmap moves—and the vendor is not optimizing for your network.
You cannot future-proof an airline. The future will not cooperate. What you can do is own the layer everything else depends on, and make adaptability the one thing you never have to buy again.