A working group, not a procurement
Alongside the product there is a policy argument, and it is deliberately separate. Nothing here needs to be bought. What it needs is a table.
The opportunity
~93% of online travel distribution in the United States sits with two companies — Expedia Group and Booking Holdings.
15–25% commission taken on each reservation, leaving the local economy permanently and never recirculating in it.
10–20% of every booking retained in-region instead, under a native channel the region owns and operates.
Today
A guest books a guesthouse in Samarkand through a foreign platform. Between 15% and 25% of that payment leaves the country. The guest relationship, the rate data and the demand signal leave with it.
The proposal
The same booking clears through a channel built and operated in the region. 10% to 20% of it stays. The operator keeps more, the traveller pays less, and the tax base grows on the same night's stay.
How the toll road was built
None of this was designed. It accumulated — one useful layer at a time, each solving a real problem, each taking a cut it never gave back. Reading it in order is the fastest way to see why the current arrangement is not a law of nature.
- 1960
SABRE goes live
American Airlines and IBM put reservations on a mainframe. Whoever controlled the terminal controlled the booking — distribution becomes a competitive weapon.
- 1988
Sixteen chains fund a switch
Each contributes $100,000 to THISCO, to translate hotel data into four incompatible GDS formats. By the mid-nineties it carries 70% of electronic hotel reservations.
- 1994
A deal collapses
Sabre and Microsoft fail to agree on building an online agency together. Microsoft builds Expedia. Sabre builds Travelocity.
- 1996
The web arrives, three times
Expedia and Travelocity launch in the United States. In Amsterdam, a small company called Bookings.nl begins listing Dutch hotels.
- 2005
A $133 million bet
Priceline buys Bookings B.V. and merges it with Active Hotels to create Booking.com — now a company worth more than $170 billion.
- 2015
The GDSs concede
Expedia buys Travelocity from Sabre and Orbitz from Travelport. Both sellers were GDS companies; both sales were admissions that they had lost the consumer.
- 2018
The switch becomes a feature
RateGain acquires DHISCO, the successor to THISCO. The layer sixteen hotel chains built to free themselves is folded into a channel manager.
- 2024
A protocol, published quietly
The Model Context Protocol gives any AI agent a standard way to call a property directly — no intermediary, no commission, no loss of the guest record.
- 2026
The next standard opens
Google announces a universal commerce protocol for agent purchasing. For the first time since 1996, the distribution layer is genuinely in play again.
The same night, three architectures
What actually stands between a room and the person who wants it. Every box was once someone's solution.
- Property
- PMS
- CRS
- Switch
- GDS
- Terminal
- Agent
- Traveller
Eight steps. Every layer takes a cut, and every layer adds a format to support.
- Property
- PMS
- Channel manager
- OTA
- Traveller
Shorter, and more expensive: 15–25% per reservation, plus the guest record.
- Property
- Open protocol
- Agent
- Traveller
The property publishes once. The agent reads it. Nothing in between rents the position.
Why the window is now, and why it closes
Distribution is being rebuilt around AI agents. Within a few years a large share of travel booking will be initiated by software acting for the traveller, and those agents will book whatever inventory is machine-readable to them. That transition is the opening: a region that becomes discoverable and bookable directly, on open protocols, does not have to buy its way into an incumbent's network. It can be found on its own terms.
The same transition is the risk. If the region arrives late, it will simply re-rent the distribution stack of the 2000s for another twenty years, at the same rates, with the same terms — rate parity, guest-data gatekeeping, and the margin leaving each time.
Proposed scope
- Establish the baseline
- Measure what the region actually pays today — commission outflow by market, by property class, and by season. The number has, to our knowledge, never been published for Central Asia.
- Define an open distribution standard
- Rate and inventory exchange that any regional property, from a Bukhara guesthouse to a Tashkent hotel, can publish to without an intermediary — and any booking agent can read.
- Pilot one corridor
- A single Silk Road route — Samarkand, Bukhara, Khiva — is enough to demonstrate retained margin against a measured baseline.
- Regionalise
- What works for one country works better for five. A single Central Asian layer is far stronger than five national ones competing for the same traveller.
What is contributed
Sixteen years in this industry, and systems built at every level of the distribution chain — the property management layer where inventory originates, the booking layer where it is sold, and the distribution layer that connects the two. All three are built and operated against a real 200-unit property rather than a slide deck. That work goes to the group at no cost, and the resulting standard should be open rather than owned.
This may also make an interesting project for the USA–Uzbekistan Business Council — a cooperation story with measurable benefit on both sides, and one that is genuinely unusual: infrastructure the region owns outright, rather than access it rents.
Get in touchCaravanserais were the Silk Road's original hospitality infrastructure — built by the region, owned by the region, and open to every traveller who arrived. The proposal is only that the digital equivalent be held the same way.