The Skift Takeaway
Skift Data + AI Summit
Data Lessons from Deploying Consumer-Facing AI
SEJAL AMIN, Chief Technology Officer, Priceline
Moderated by Sean O’Neill, Senior Hospitality Editor, Skift
THE ARGUMENT
Priceline relaunched its AI assistant Penny around a single design principle: go from search to book in one unbroken flow. The technical challenge was real: include a network of agents orchestrating in real time, with a map interface that updates dynamically. But the more consequential bet is on preference data. Penny now combines the behavioral data that Priceline has stored for years with preference data it collects from users — and lets them edit it. The cost question is unresolved: Priceline is shipping features faster than ever, but Amin said new features are not automatically cost-effective, and routing requests to the right models at lower cost is now driving the roadmap.
THE EVIDENCE
- Penny users convert at a higher rate than non-Penny users. — Amin
- A trip to Montreal would require Penny to summon 10-12 agents, orchestrated in a single flow invisible to the user.
- “Just because we have new features, it doesn’t necessarily mean they’re all cost effective.” Token spend and model routing are now the central architectural debates.
- “We’ve got to figure out how to optimize our architecture, the decisions we’re making there, and figure out how to route requests to all of the right places at a lower cost.”
THE SO WHAT
Priceline’s transparency play — show users their data, let them correct it — is a direct answer to the trust problem that has slowed AI adoption in consumer travel. If that compounds into higher conversion over time, it becomes a structural advantage over platforms that treat preference data as a black box. The harder near-term problem is cost: deploying the best model for every task is not sustainable at scale, and any travel company building consumer-facing AI without a model-routing strategy is accumulating a cost problem it has not yet priced.
Access all the Skift Takeaways from this event, as well as full session videos, presentation documents, and more editorial coverage at our Skift Data + AI Summit attendees hub.