Travel: Orchestration, Not Experiments

Travel companies spent the last two years building chatbots that could answer “Where should I go in July.” In 2026, the harder job is getting an artificial intelligence system to actually search flights, apply loyalty status, hold a room, add a transfer, and fix the trip when one leg breaks. (skift.com) That shift has a name inside the industry: orchestration. Instead of one giant assistant doing everything, companies are wiring together specialized software agents that each handle a narrow task across booking, distribution, and customer service systems. (skift.com) Travel is a messy place to try this because one vacation can touch an airline reservation system, a hotel supplier, a car rental feed, a payment processor, and a loyalty database in a single checkout. Skift’s April 8 report says the winners will be the companies that make those old systems work together at scale, not the ones running flashy artificial intelligence demos on top. (skift.com) That is why the new race is less about writing prettier answers and more about building plumbing. If an agent can suggest a better fare but cannot confirm inventory, enforce fare rules, or process a change after a delay, it is just a smarter search box. (skift.com) The catch is that a booking agent that can act is also a tool that can be abused. TravelMole reported on April 8 that DataDome is warning travel brands that agentic booking flows can raise fraud risk because the same automation that helps shoppers can also help attackers move faster. (travelmole.com) DataDome says the old security model was mostly about spotting bots and blocking them. In agentic commerce, the harder question is whether the visitor is a legitimate software agent acting for a real traveler with permission to search, book, change, or refund. (travelmole.com) That distinction matters in travel because prices and availability change minute by minute. A malicious agent can test cards, scrape fares, drain loyalty accounts, or exploit refund loops at machine speed if the booking pipeline trusts automation without checking identity and intent in real time. (travelmole.com) (sabre.com) TravelMole says DataDome estimates more than 90% of travel sites are not fully protected against basic automated threats. Sabre has made the same point from the supplier side, arguing that autonomous systems that can search, price, book, change, and refund in one loop need identity safeguards, behavior monitoring, and policy checks built into every step. (travelmole.com) (sabre.com) So the next version of travel artificial intelligence will look less like a concierge and more like air traffic control. The valuable companies will be the ones that can coordinate many systems at once while proving, in milliseconds, which agents are trusted, what they are allowed to do, and who pays when something goes wrong. (skift.com) (travelmole.com)