An independent guesthouse owner and a single tour guide face the same problem a hotel chain solved with a revenue team: what should this weekend cost, and is the listing copy doing any work at all?
The evidence side of that is already built. A provider dashboard computes, live from real bookings for the listings that provider owns, bookings and revenue over time at day, week, or month granularity, a cancellation rate, a capacity-utilization rate, and a top-listings drill-down with per-listing revenue and cancellation figures. It is deliberately honest about its own limits: utilization is measured as guests booked against listing capacity, because there is no per-date room or seat inventory table to compute true night-by-night occupancy against, and forecasting numbers were explicitly left out of that work rather than faked.
The copilot is the layer that reads those numbers and answers a question instead of drawing a chart. Designed behaviour: look at a listing's utilization and cancellation pattern, look at comparable inventory in the same destination, and return a suggested rate for a given date range together with the reasoning and the figures it used — so a host can disagree with it on the evidence rather than on vibes. The same engine drafts listing titles, descriptions, and amenity summaries from the structured fields a host has already filled in.
Two rules shape it. First, it advises; it never silently repriced anything. A recommendation is a card a host accepts or rejects, and accepting it writes through the ordinary listing-update mutation with the ordinary audit entry. Second, it can only ever read the provider's own data — the analytics query already enforces a strict single-provider ownership check and never crosses providers, and the copilot inherits that boundary rather than being trusted to respect it.
None of the pricing or drafting is built yet. The analytics substrate it needs, the listing-update path it would write through, and the audit trail that would record an accepted recommendation all exist today.
Illustrative scenario. Provider analytics are live; AI pricing guidance and listing-content drafting are designed and ticketed, not shipped. No real rates, bookings, or hosts are implied. This story is narrative-only for now — there is no interactive "Do it yourself" walkthrough because the copilot is still being built.
Ready to make this your story?
