Short-stay revenue lives in the gap between the rate a host sets and the rate the market would have paid that night. A stale rate rarely shows up as a lost booking on a single night — it shows up as a quiet gap on a year-end P&L, across many nights, against an unseen calendar.
AI pricing closes that gap without replacing the operator. The model reads your calendar, your house rules, your historical occupancy, and the live comp set, then proposes a curve. You approve a band. Decisions that change the published rate stay with the operator of record.
What the model actually does
Every hour the pricing job looks at the next 90 nights, classifies each one (weekend, holiday shoulder, midweek shoulder, low-demand), and proposes a rate within the band you approved. It surfaces the deltas so an operator can scan the diff and mark which nights to publish, which to hold.
It does not invent events. It reads public calendars and the reservations it can already see. When it cannot see a competitor (blocked calendar, opaque inventory), it says so — silence is not a price recommendation.
What it does not do
It does not set revenue-share floors. It does not pretend to know about conferences your city has not announced. It does not commit you to a rate — every publish is a single click.