The trip began with a single sentence typed into a chat window at 11pm: “Tokyo to Reykjavík over ten days, two stops, nothing before 9am, window seats, vegetarian.” Forty seconds later, the agent had drafted a full itinerary, flagged three weather risks, and quietly held two flights at fare-lock prices while waiting for approval.
What used to be a travel agent's week of work is now a background task. The interesting part is not the speed. It is the way the agent keeps working after you stop paying attention.
From assistants to agents
A travel chatbot answers questions. An agent takes actions: it watches the price of your held seats, swaps a hotel when a better-rated one drops into budget, and rebooks a missed connection before you reach the gate. Across the trip, our agent intervened seven times. Three of those interventions we never noticed until reviewing the log afterward.
The interesting part is not the speed. It is the way the agent keeps working after you stop paying attention.
Where it still gets it wrong
Edge cases are the agent's weakness. A festival in Akureyri shifted the cheapest hotel into a hostel with a noise complaint history. The agent could not tell. A human travel desk would have flagged it on sight. We are not in a world where the agent replaces local expertise — we are in a world where local expertise becomes a premium layer on top of an agent that already did the boring 80%.
What this changes for destinations
If most itineraries are now drafted by agents, destinations need to be machine-readable. Open APIs for opening hours, accessibility, capacity, and pricing become as important as a beautiful website. The destinations that win the next decade will be the ones an agent can confidently recommend at 11pm to someone who has never heard of them.
About Sophia Goehner
Sophia covers applied AI and the changing shape of service work. Based in Ithaca, New York.
