Nearly every AI product in travel is built for the person taking the trip. Planners, itinerary generators, chat concierges, personalisation engines, inspiration feeds.
All of it assumes the bottleneck is the traveller's decision. On the data I have, it is not.
What the data says the constraint is
More than 36,100 business-to-business wholesale hotel bookings, one Gulf desk selling to travel agencies, more than 7,340 properties across more than 110 countries. Two facts out of that file decide the argument.
More than 60% of bookings were created outside office hours. The five busiest hours of the day all fall after most desks have closed. Finding 01.
The median gap between booking and travel was nine days, and more than 47% travelled inside a week. Finding 02.
Put those together and the picture is not a traveller who cannot decide. It is a traveller who has already decided, at 21:00, for a trip nine days away — and a desk that is closed.
The customer was never the slow part.
The thing consumer AI cannot reach
There is a reason the industry's AI attention stops at the traveller, and it is not stupidity. It is that the traveller-facing layer is the only part of travel that is easy to reach.
Behind the desk sits the part nobody demos: assembling a quotation that is complete, priced, branded and comparable, from sources that were never designed to be read together, with the owner's margin applied and the cancellation terms attached. That work is not glamorous, it is not visible from the outside, and it consumes the majority of a desk's productive hours.
It is also the work that decides whether you win the booking. On a nine-day median, the agency whose complete answer exists first usually wins — not the one whose answer is best.
Why this is not an argument for replacing agents
I want to be exact here, because the loose version of this argument is wrong and I keep hearing it.
AI is not replacing travel agents. The judgement in this job — what to sell, at what margin, to which client, and when to break your own rule because you know the family — is the entire value of the person. It is also the part with real downside when it goes wrong, and there is no version of this I would hand over.
What AI replaces is the assembly. The gathering, the formatting, the repetition, the re-typing of the same information into a shape a client can read. That is the work that fills the day and produces nothing a customer values in itself.
An agency that automates the assembly and keeps the judgement gets faster without getting worse. One that automates the judgement gets a liability with a friendly interface — particularly if it ignores cancellation policy, which on this data varies by close to three times depending only on the destination.
What I would build if I were starting now
The same thing I did build. A message from a client becomes a complete, priced, branded quotation while the client is still in the conversation — and the owner approves it before it goes anywhere.
Not because approval is a compliance box. Because the desk's judgement is the product, and the assembly was never the part worth protecting.
I spent 17 years inside travel operations before I wrote a line of this. That is the only reason I know where the hours go. The industry does not have an AI problem. It has an assembly problem wearing an AI costume, and it keeps buying answers for the wrong end of the transaction.
The demand arrives at 21:00. Build for the person who has to answer it.
Source: the Gulf Outbound Hotel Index, version 1.0 — more than 36,100 B2B wholesale hotel bookings from one Gulf desk. Findings and full tables · Methodology · CC BY 4.0 · 10.5281/zenodo.21796038