Two different clocks
Busch's core distinction: hotels are running two AI adoption tracks that look similar and are not.
- Internal AI — forecasting, pricing, ops copilots. Optional. Vendor-paced. The hotel decides the timeline.
- AI-native distribution — being bookable by AI shopping assistants. Not optional. Market-paced. The timeline is set by consumer behaviour, not the hotel.
Treating the second one like the first — as a project to schedule for next quarter — is the mistake being made across the industry right now.
Today's distribution assumes humans
Current channel management assumes a human on the other end reads a description, compares a few results, and clicks. So hotels push static descriptions, static rates, and wait to be scraped and ranked. When machines are querying inventory in real time, static-push distribution simply does not answer the question.
What AI-native distribution actually requires
- Machine-readable facts — amenities, availability, room attributes, cancellation terms, in a structured schema
- Live data connections — assistants query in real time; there is no batch job that saves you
- Dynamic pricing exposed via API — the price the machine gets has to be the price the guest gets
- Direct booking acceptance from agents — the transaction closes in the assistant, not on your site
- Merchant-of-record readiness — chargebacks, disputes, refunds land with the hotel, not the OTA
The operational implication
Moving to AI-native distribution shifts merchant-of-record burden onto hotel teams that have leaned on OTAs for exactly that work. The staff-cost model changes. The dispute-handling process changes. The finance workflow changes.
What to do this quarter
- Stop treating distribution readiness as internal-AI's little sibling
- Audit whether your existing channels can respond to real-time machine queries
- Pressure your PMS/RMS/CRS vendors on their AI-native roadmap dates
- Prepare ops teams for merchant-of-record duties they have not owned before
- Assume the first AI-native distribution wave lands in months, not years