A CDO walks a commercial leadership team through what AI could do for the organization. Agent copilots for analysts. Real-time demand signals. Natural language queries across the data landscape. Board-ready, ambitious, on-trend.

Twenty minutes in, a commercial officer cuts across it.

“We’ve been trying to get our numbers to agree for years. Finance has their version. Revenue Management has another. Every quarter we lose a week reconciling before our board meeting. Now we’re going to put AI on top of it?”

The room doesn’t argue. The CDO doesn’t either. They’ve heard the frustration before. What they don’t have is a board slide that reads “fix the foundation.”

I’ve been in some version of this meeting for a long time. Twenty years ago the technology on the table was the enterprise data warehouse. A decade ago it was cloud. Each of those shifts was pitched as the moment the data problem would finally get solved. Each one mostly didn’t.

Not because the technology was wrong. It worked. The problem was that each of those projects was measured by technical milestones on a vague mission.

“We went live.”

“We migrated.”

“We’ve modernized!”

Rarely did we track whether the business gets better answers, faster. And the oxygen that should have gone into maturing the data around an aligned and empowered business got consumed by software upgrades and platform migrations. The technology always arrived. The insights rarely did.

AI is the biggest of these shifts, amplifying the same organizational patterns but in such a way that these latent data problems might finally break us. A report you don’t trust is a problem you can live with. An agent acting on numbers that don’t agree is a problem that scales fast.

We think the problem is that we don’t have the data. But the data is already there. It’s been there for years. The problem isn’t scarcity.

What an airline captures in a day

Every shopping session. Every fare filing. Every booking, exchange and cancelation. Every ticket issued and refund processed. Every boarding pass scanned, bag tagged and ancillary purchased. Every tier qualification and award redemption. Every email opened, every mobile session, every web conversion funnel entered and exited.

All of it captured. All of it timestamped. All of it stored.

An airline’s commercial operation is one of the most data-rich environments in any retail industry. Nothing moves through the system without leaving a trail. The data on a Tuesday afternoon in April is more granular than most retailers produce in a month. And customer identity resolution? Few purchases require checking a legal name, gender and birthday more times than traveling by air.

So when a commercial leader asks whether we have enough data, the honest answer is yes—many multiples over. The harder question, the one nobody quite wants to ask out loud, is why can’t we get to it when we need it.

The fragmentation is architectural

Airlines didn’t wake up one day and decide to scatter their commercial data across a dozen systems. It happened the way airlines built themselves—one system at a time—and each new system quietly redefined what a customer was and what counted as revenue.

The departure control system first. Then the PSS. Revenue accounting arrived with its own definitions—what was sold, what can be recognized, proration. Reconciliation just across these three systems alone has been a cottage industry for decades. So many airlines just don’t even try.

Loyalty arrived next, homegrown in the early 80s as a mainframe database. Loyalty introduced a different definition of the customer—not the passenger on a PNR but a member with a history.

Digital layered on top of all of it. Sessions, carts, funnels, events. Revenue in the digital stack is an e-commerce transaction with its own attribution rules. The customer is a cookie, an email, a device—authenticated sometimes, anonymous often. A whole new vocabulary for revenue and customer, built on top of everything underneath it that had never been reconciled.

Somewhere along the way a CDP got procured to pull it all together—and today it pulls in some of it, sometimes.

The result is the fragmentation everybody lives with. None of this is one person’s fault. But it is everyone’s job to fix it.

Every system made reasonable choices inside its own scope. What was missing—what is still missing at most carriers—is a layer that exists above any individual vendor. A layer where the business rules of the airline itself are defined, governed and served. What constitutes an order. How is revenue recognized. Who is our customer and who are they today.

Without it, every question that moves between systems becomes a reconciliation project.

What the decisions actually need

Inventory managers need booked and forecasted demand, authorization levels, all with point-in-time history as the flight fills. All of it exists. Stitching it together without the latency decision makers demand is the problem.

Merchandising needs to know what the customer bought, what they considered buying, what else they might buy and what the airline has to offer. This data is all captured, too. The systems with it let most of it evaporate.

Flight profitability needs revenue and cost pulled together on the same flight, at the same level of granularity, with the same definitions. Revenue lives across the PSS, revenue accounting, merchandising, settlement and sometimes loyalty. Cost lives in a whole other world between the operation and accounting.

The pattern repeats for every commercial decision. As carriers shift from leg-based to O&D revenue management, analysts are being asked to reason about trips that cross their network, not just the legs in front of them. The data to do that work needs unification. Most analysts have never seen it in a form they could act on. They just assume the automated tools they’ve bought have it and will.

The inputs exist. The context doesn’t.

Why AI is the test

For twenty years, this fragmentation has been inconvenient. Analysts work around it. Commercial teams staff up to reconcile it. Every airline has someone—usually several people—who know where the real numbers live and how to piece them together. The cost is high, but it’s hidden inside headcount and process.

AI doesn’t let us hide it anymore. For a while now, Tableau has been making bad data believable by making it pretty. Now, LLMs are making bad data believable by making it sound real. A chatbot reading from three systems with three definitions of revenue will give a confident answer that is quietly, structurally wrong. An agent acting on that foundation doesn’t just produce a bad report. It produces a bad decision, at scale, with the air of authority that conversational interfaces naturally carry.

The airlines that deploy AI on top of fragmented data are not going to get AI. They’re going to get bad decisions in a confident voice.

Own the layer

The layer that’s missing isn’t a product. It’s not a vendor’s fast new data lake. It’s not a connector or a new cloud play.

It’s an internally-owned layer where the airline’s own rules about order, customer, revenue and flight are codified once, governed by the business, and served everywhere they’re needed. Some people call it the silver layer. But however it shows up on an architecture diagram, it’s the valuable layer between raw system output and everything your commercial team wants to do with it. All the data. All the history. All the definitions in a single model. All joinable and available to the enterprise.

That layer is how a commercial leader gets a trustworthy answer without a six-week reconciliation. It’s how a new merchandising platform plugs in without another point-to-point integration. It’s how analysts start reasoning in true O&D instead of flights. And when the time comes, it’s how an AI agent gets context rich enough to actually help.

Success in this latest shift won’t be decided by who has the most data. Every airline already has enough. It’ll be decided by who treats their data as theirs to command, not a vendor’s to mediate.

The right partners to help us with this shift will be the ones that expect our data foundations to be sound and want to work with us to increase our agency with it—not transform it into yet another version we have to rent.

When you find yourself thinking you wish you had the data, remember the data is already there. What you need is your data in a layer that you own and that works for you.