Booking’s 5 AI Bets vs Expedia’s Agent Shift

OTAs are not dabbling in AI. They are portfolio-managing it.

Skift (Sept 14, 2026) details Booking Holdings’ five AI bets: trip tools at Priceline, Agoda, and Booking.com, plus two internal startups. Priceline’s Penny has shown early CSAT, engagement, and conversion gains—on a small sample. Expedia, meanwhile, dropped its end-to-end Romie approach in favor of specialized multi-agents.

Two giants. Two strategies. One clear message for hotels: the way trips get planned is being rewritten in public.

The story

One company, five experiments. That is how a scaled travel platform hedges model risk, UX risk, and brand risk. Brand-level trip tools keep learning close to the consumer at Priceline, Agoda, and Booking.com. Internal startups keep optionality when core apps move slower than the lab.

Penny’s early lift on CSAT, engagement, and conversion is directionally encouraging—and honestly framed as a small sample. That caveat is useful. Travel AI is full of demos that die in production. Early metrics with honesty beat silent rollouts that hide failure.

Expedia’s pivot is equally instructive. An end-to-end assistant (Romie) gave way to specialized multi-agents. The industry takeaway writes itself: monolithic “do everything” travel agents may lose to focused agents that cooperate on pieces of the trip. Skift’s coverage also flags why Agoda’s experiment is next to watch inside Booking’s portfolio—another reminder that this is a multi-brand race, not a single product launch.

Stakes for hotels

Hotels do not control OTA AI roadmaps. They live with the outcomes. When Booking and Expedia reshape how travelers plan and book, hotel visibility, content quality, packaging, and direct-channel competitiveness all move.

Implications for USA hotels and brands:

• Content readiness — AI trip tools will surface properties that are structured, accurate, and differentiated—not just discounted.
• Direct channel pressure — OTA AI that converts better raises the bar for brand.com, apps, and call centers.
• Partner strategy — Know which Booking brand’s AI your demand flows through, and what that agent optimizes for.
• Architecture lesson — Specialized multi-agents (Expedia’s direction) may change how recommendations and add-ons get assembled around your hotel.

What it means

Booking is diversifying bets. Expedia is specializing agents. Both paths admit the same truth: the first generation of all-in-one travel AI was a hypothesis, not a finished product. Hotels that treat OTA AI as “someone else’s roadmap” will wake up to share shifts they never modeled.

Commercial teams should brief owners on both plays: five experiments at Booking Holdings, and a multi-agent pivot at Expedia. Then stress-test direct offers against agent-assisted OTA flows before the gap widens.

Takeaway

Watch the experiments, not the slogans. Soft CTA: brief your commercial team on which OTA AI surfaces your inventory—tighten content and direct offers before specialized agents get better at closing the trip without you in the conversation.