IBS Group Spins Out Naviq: Travel Tech Finally Gets Its AI-Native Moment

The travel industry has spent two decades bolting AI onto mainframes that were old when dial-up still screamed. IBS Group just admitted what everyone in the back office already knew: you can’t retrofit intelligence into a codebase that predates the iPhone. Their answer is Naviq, a standalone, AI-first company carved out of IBS’s own guts. That’s not a pivot. It’s a bankruptcy declaration for the old stack — and it’s the most honest move a travel-tech incumbent has made in years.

Why the Spinout Matters More Than the Press Release

Press releases love words like “synergy” and “ecosystem.” Naviq’s launch materials are no exception. But strip the fluff and the signal is clear: IBS realized its core platform — the one running reservations, inventory, and pricing for airlines and hotel chains — cannot become AI-native without breaking every SLA it signed. So they built a clean room, hired a team that doesn’t know what a green screen looks like, and gave them a mandate: rebuild the decision layer from scratch.

That decision layer is where the money lives. Not in the PNR. Not in the CRS. In the micro-decisions that happen millions of times a day: which fare to surface, which ancillary to bundle, which rebooking path to offer when a connection breaks. Legacy systems make those decisions with rules engines that haven’t seen a gradient descent since Y2K prep. Naviq bets that a model trained on real-time demand signals, competitor moves, and operational constraints will beat the rule set every time. History says they’re right — if they can get the data out of the mainframe without melting the pipes.

The Data Plumbing Problem Nobody Talks About

Here’s the part the launch blog skips: Naviq’s models are only as good as the feature store feeding them. IBS sits on petabytes of booking histories, no-show patterns, disruption recoveries, and ancillary attachment rates. That data lives in formats that range from modern event streams to VSAM files that still think in 80-column records. Extracting, cleaning, and versioning that corpus without disrupting live traffic is a DevOps nightmare dressed as a data-science project.

I’ve seen three “AI-first” travel startups choke on exactly this. They build beautiful TensorFlow pipelines, then spend eighteen months negotiating change-control boards to add a single column to a feed that touches revenue accounting. Naviq’s advantage is structural: they share a cap table with the data owner. That should shortcut the politics. Should. The moment a quarterly earnings call asks why the new model dipped revenue by 0.3%, the politics return. The real test isn’t model accuracy. It’s whether Naviq’s leadership can say “roll it back” without getting overruled by the parent’s CFO.

What This Means for the Rest of the Stack

Amadeus, Sabre, Travelport — they’re all running the same playbook: acquire an AI lab, publish a white paper, integrate a chatbot into the agent desktop, call it transformation. Naviq’s spinout model is different. It says the core platform is legacy, and the future gets its own P&L, its own hiring plan, its own technical debt budget. That’s the model every incumbent should copy but won’t, because it requires telling Wall Street that the cash cow is depreciating faster than the depreciation schedule admits.

For airlines and hotel groups, the signal is simpler: your vendor roadmap just forked. The old roadmap promises stability. The new one promises intelligence. You can’t buy both from the same codebase anymore. Procurement teams need to start writing RFPs that separate “system of record” from “system of decision” — and stop pretending a single vendor wins both.

The Bet Worth Watching

Naviq will live or die on one metric: time from signal to action. When a competitor drops a flash fare, when a volcano closes an airspace, when a crew timeout cascades through a bank — how many seconds until the model reprices, rebundles, and reoffers across every channel? Legacy systems measure that in batch cycles. Naviq’s architecture targets sub-second. If they hit it, they redefine what a travel platform is. If they miss, they’re just another analytics dashboard with a better logo.

IBS gave them the data, the capital, and the customer access. That’s more than most AI travel startups get. It’s also exactly what makes the next twelve months dangerous. Comfort breeds slippage. The team that ships the first production model at 200ms latency — and survives the change-control board — writes the next chapter. Everyone else is still writing PowerPoints.

Source: original report