Citi Says AI Companies Should Pay for Agent Travel Booking Errors

What happened

Citi has argued that AI companies, not hotels and airlines, should pay when the travel agents they deploy book the wrong itinerary. The position, reported by Skift, responds to a live question in travel: AI agents are already making bookings, they sometimes get them wrong, and the sector has not agreed who absorbs the cost of those errors.

According to the reporting, Citi's view would place financial responsibility with the companies that build and operate the agents. That would relieve suppliers such as hotels and airlines from covering mistakes they did not make. The same reporting notes a trade-off: a liability model of this kind could slow how quickly agents are able to complete transactions.

Why it matters

Agentic booking is moving from demonstration to everyday use, and the question of who pays for errors is becoming a question of operating design. Whoever carries the cost of a wrong booking will shape how cautiously agents are built, how much human confirmation is inserted into the flow, and how quickly a failed transaction is detected and reversed. For leaders in digital transformation, this is a question about accountability architecture as much as technology.

For customers, the moment an agent gets a trip wrong is the moment trust is either repaired or lost. A clear route to correction, with a named party responsible for putting things right, will matter as much to the traveller as the quality of the original booking. Organisations deploying agents will need to plan for that recovery experience, not only for the happy path.

The Renascence take

Most of the debate is framed as a fight over who writes the cheque. The more useful question is who the traveller believes they are dealing with when something goes wrong.

Liability rules will not fix a booking experience that leaves people unsure who is accountable. Customers who delegate a decision to an agent are not thinking about the supply chain behind it, so a failure reads as the brand's failure, whoever ends up paying. The behavioural principle at play is that trust is rebuilt at the point of error, not at the point of purchase. Operators should design the recovery path now: a visible owner, a fast fix and a clear explanation, whatever the eventual cost split turns out to be.