Programmable Commerce for Hospitality

Ask a night auditor how the property got paid for last night's arrivals and the honest answer is: four different ways, on four different timelines, with four different people to call when the numbers do not match. The guest who booked on an OTA under the merchant model was charged weeks ago by the OTA; the hotel will see a virtual card after checkout. The guest on an agency reservation paid the front desk; the commission invoice arrives next month. The prepaid non-refundable guest was charged at booking, card not present, and can dispute it for four months. The guest who used an installment plan paid a lender, who paid the hotel less a fee the hotel did not negotiate.

This article is a technical and business analysis of that stack and of what replaces it. The vision is programmable commerce; the mechanism is programmable settlement: terms travel with the transaction and are enforced by protocol, not by trust in an intermediary. The argument is not that OTAs pay slowly out of malice — they built the demand and still deliver it — but that a gatekeeper who controls the payment flow controls the terms for every other participant, including the ones that have not been invented yet.

Three things are worth reading across the rows. First, the party holding the money is never the party bearing the operational risk of the stay. Second, the fee stack compounds: on a merchant-model reservation the hotel pays commission and then pays its own processor to charge the OTA's virtual card, a card fee on money that was already the hotel's. Third, every incumbent flow adds a reconciliation job, and reconciliation is where commission leakage hides (see Dark Patterns in OTA Extranets).

The card fee is the easiest number to source. The standard published online rate for a US card transaction through a major processor is 2.9% plus $0.30. On a $129 night that is $4.04; the hotel keeps $124.96 and sees the money in two to three business days. Under a merchant-model OTA reservation at 18% commission the same night nets $101.82 after commission and the card fee on the virtual card, and the payout arrives 30–60 days later.

ADAPT's protocol-interfaces report ran that arithmetic at the scale of the Exchange Building pilot. Illustration, with its assumptions stated: 200+ units generating roughly 10,000 bookings a year at a $129 average; card fees of about $40,000 a year; OTA commission at 18% of about $232,000 a year; combined, roughly $272,000 a year that leaves the property under the incumbent stack. It is a model, not a measured result, and the booking count is an assumption, but the shape is what matters: the commission line is five to six times the card line, and both are set by someone other than the hotel.

The less visible cost is latency and its cousin, uncertainty. A property that does not know which of its virtual cards will fund, or whether a commission invoice includes the stay that cancelled, carries working capital against that ambiguity. Small operators feel it as a line of credit drawn in the slow season; larger ones feel it as a finance team.

Programmable settlement replaces the closed set of flows with one open one. The terms of the booking — rate, cancellation window, deposit, dispute reserve, commission or advisor split, payout schedule — are encoded in the transaction when the AI agent confirms it. When a condition is met, the corresponding transfer executes. A cancellation inside the free window refunds automatically because the window was a term, not a policy page. A no-show charge posts because the guarantee was a term. A dispute freezes only the reserve because the reserve was a term.

The wire protocol ADAPT has been tracking for this is x402, an open, HTTP-native payment protocol for machine-to-machine settlement, published in 2025. It revives the HTTP 402 'Payment Required' status: a property's endpoint answers a booking request with 402 and the price and terms; the agent retries with a payment header; the server verifies and confirms. Settlement finality is measured in seconds, and the protocol itself charges no fee. Google's Agent Payments Protocol (AP2, September 2025) and OpenAI's Agentic Commerce Protocol with Stripe (September 2025) address the same layer from the platform side, and Google's Universal Commerce Protocol (January 2026) frames the commerce flow above it. ADAPT's position is that a property should be able to accept whichever of these an agent presents, and that the terms should be the hotel's.

Three properties follow. Instant clearing: the hotel's share is available when the booking is confirmed, not after a batch cycle. Enforcement by protocol: the cancellation and refund logic runs from the encoded terms, so there is nothing for a support desk to override and nothing for a finance team to reconcile. Open participation: because the terms are a readable part of the transaction, any third party that can read them can offer to finance, insure, or guarantee them, which is the subject of the rest of this article.

Once the terms of a booking are legible to machines, a reservation becomes something a lender can underwrite one at a time. Today that market is closed: the only installment option a guest sees is the one the OTA has integrated, on the OTA's economics. Programmable settlement opens it to the three kinds of local participant ADAPT's manifesto names.

Credit unions, per booking

A local credit union already knows how to make a small, short-term consumer loan; what it lacks is the point of sale. With terms attached to the transaction, a credit union can offer a member per-booking credit — the funds are conditional on the stay and the cancellation logic is inherited — and be repaid on a schedule the member chooses. In the United States credit unions are member-owned cooperatives supervised by the NCUA, which is what makes them a natural fit for a zone-bounded product: the members are the residents and the visitors' hosts.

Community development funds

Community Development Financial Institutions, a US Treasury certification, exist to lend where conventional capital does not. A CDFI in a tourism-dependent county can underwrite travel loans for visitors to its zone the way it underwrites inventory loans for the businesses those visitors patronize, and the protocol gives it the same conditional-payment rails as a national BNPL provider. The economic-development case — a booking funded locally, spent locally, and settled locally — is the same case DMOs already make for direct booking.

Tourism Development Fund zones

The manifesto's third participant is the Tourism Development Fund zone: a geography, designated with the local DMO, inside which the protocol's rule-set can carry a small local commission (1–2% for an AI-empowered local travel advisor), a DMO contribution on event blocks, and zone-specific terms for financing and arbitration. The zone is what turns a payment protocol into a local financial services layer: the credit union, the CDFI, the advisor, and the arbiter all participate under rules the community set rather than terms a platform imposed. Consumer protection travels with the money. A per-booking loan is a loan, subject to disclosure rules, and a bonded guest's deposit is held by the protocol, not by the lender; nothing here relaxes those obligations, it makes them enforceable at the point where funds move.

The most novel participant is the one ADAPT's manifesto calls the certified arbiter as initial financier. The local arbiter who would hear any dispute on a reservation also funds it: the arbiter pays the property immediately at booking, collects from the guest on agreed terms, and earns a financing spread, plus the arbitration fee if a dispute arises.

The incentive logic is straightforward. An arbiter who rules carelessly loses money in either direction. A wrongful decision for the property leaves the arbiter holding a receivable from a guest who will not pay it and who can still dispute the charge with their card issuer. A wrongful decision for the guest leaves the arbiter absorbing a refund the property's reserve may not cover. An arbiter who resolves accurately and quickly earns the spread on every clean booking and the fee on every contested one. Lender, guarantor, and judge merge into one local actor with exposure on both sides. ADAPT's working notes extend the idea: the same arbiter can underwrite the guest trust deposit for travelers in the zone, so one actor holds three aligned roles.

The risks are real, and the working group should name them before critics do. First, the founding framework lists 'a current financial relationship with a party to the dispute' as grounds for arbiter ineligibility, and a financing arbiter has one with both parties. The defensible answer is that the exposure is symmetric and disclosed at booking, and that a financing arbiter recuses from the decision itself while remaining the guarantor: a peer arbiter decides, the financier pays. Second, an arbiter who is also a creditor may prefer outcomes that keep receivables collectible; outcome-neutral fees and the appeal panel are the checks. Third, concentration: a single arbiter-financier in a small market is a single point of failure for both fairness and liquidity, which argues for a minimum of two certified arbiters in any zone before financing is enabled. Fourth, licensing: in most jurisdictions extending consumer credit requires a license an arbiter may not hold, which is why the credit-union and CDFI variants are likely to arrive first.

The abstract promised revenue projections for the new financial services layer. What follows is an illustration with stated assumptions, not a forecast. Take a mid-size destination with 2.0 million room-nights a year at a $150 average rate and a 2.5-night average stay: 800,000 bookings, $375 each, $300M in room revenue. Assume that in the third year of a zone, 5% of that volume settles on ADAPT rails: 40,000 bookings and $15M.

On these assumptions the local financial services layer earns roughly $280,000–$360,000 a year on $15M of flow — about 1.9% to 2.4% of it — and nearly all of it stays in the zone; the arbitration line is a distribution of the network fee, not an addition to it. The hotel's all-in cost remains inside ADAPT's under-8% target with room for hosting and support. Two honest observations from the arithmetic: at a 5% dispute rate the 0.5% network-fee floor does not cover arbiter fees, so the framework's range only works if disputes fall as listings and agents improve; and change any assumption and the totals move. The point of the exercise is the ratio, roughly an order of magnitude below the incumbent stack, not the totals. ADAPT will publish measured figures from the Exchange Building pilot when there are enough settled bookings to report without disclosing proprietary data.

ADAPT proposes that settlement terms belong to the hotel and the guest, that they travel with the transaction, and that the layer beneath every booking be open to any lender, underwriter, or arbiter who will serve those two parties on published rules. Concretely, that means three pieces of work.

1. A settlement-terms specification that encodes rate, cancellation, deposit, dispute reserve, payout schedule, and zone splits in a form any conforming payment protocol — x402, AP2, ACP — can carry.
2. A financing-participant standard: how a credit union, CDFI, bonding provider, or arbiter-financier reads the terms, offers credit or a guarantee, and is repaid, including disclosure and recusal rules.
3. A pilot at the Exchange Building, instrumented for settlement latency, dispute rate, and cost per booking, with results published as measured figures rather than illustrations.

The register lists this article as a proposed joint session of the Dispute Resolution Working Group and a new finance working group, and the pairing is deliberate: the arbiter-as-financier model cannot be specified by either group alone. Payment and settlement specialists, credit-union and CDFI lenders, DMOs designating their first Tourism Development Fund zones, and hotel advisors who model distribution cost for owners are the people the session needs. The property side of the same story — how a PMS publishes terms and consumes settlement events — is the subject of The Protocol-Native PMS.