“5 days, somewhere green, two adults
and a 7 year old, under 9000”
and a 7 year old, under 9000”
An actual enquiry shape · received most days on WhatsApp
They message it on WhatsApp every day — "5 days, somewhere green, two adults and a child, under 9000." Then a human turns that into an itinerary over the next two hours. This is that, in ninety seconds, on top of the booking engine we already run.
This is the argument for building it here rather than watching someone else build it.
HolidayMakers.com already has inventory, pricing, payments and apps in both stores. The planner is a front end onto systems we run, not a new company.
Every enquiry that lands on WhatsApp is a human reading a paragraph and composing a trip. That is the exact task, done at human speed and human cost.
HolidayMakers.ai is already registered and held. If we do not build on it, its only value is stopping someone else from using our name.
| STAGE | TODAY, ON WHATSAPP | WITH THE PLANNER |
|---|---|---|
| Customer describes the trip | Types a paragraph. Waits. | Types the same paragraph. |
| First response | 20 minutes to 3 hours, depending on staffing and time of day | Under 30 seconds, at 2am, in Ramadan, on a Friday |
| Options offered | Usually one, because building a second takes as long as the first | Three, priced, with the trade-offs named |
| Customer changes their mind | Another round trip. Often where the enquiry dies | Instant. Recomposes and reprices in front of them |
| Staff involvement | Full itinerary built by hand before we know if they are serious | Staff step in once there is genuine intent |
| What we learn | Nothing structured. It sits in a chat log | Every constraint and rejection, as data we can price against |
The honest version: this does not replace anyone. It removes the unpaid first hour that agents currently spend on enquiries that were never going to convert, so they spend their time on the ones that will.
Composed from our own inventory and our own contracted rates — not from a model's memory.
OPTION 1
Illustrative. Prices and availability in the real product come from our systems, never from the model.
A support bot answers questions about an existing booking. This composes and prices a product. Different job, different value, and customers can tell the difference immediately.
It proposes. A human confirms availability and takes payment. Letting a model commit real inventory at real prices is a risk with no upside at our size.
It handles the first pass. Complex multi-city, group and corporate work still goes to people, and always will.
Model calls cost money per conversation. That cost is real and must be budgeted against conversion, not waved away. Phase 1 exists to measure exactly this.
Six weeks, one destination, WhatsApp only, measured against how the same enquiries perform today. If it does not beat the manual process, we stop and we have learned something cheaply.