A multi-property luxury hospitality group came to us with a familiar problem: every reservation still ran through people on the phone. Calls spiked at the worst times, after-hours enquiries slipped away, and scaling the front desk meant scaling payroll.
They switched their inbound reservations line to HeyKoala's enterprise voice agent for hospitality. Here is the brief we took, and what the first full month produced.
What we promised
- Answer every reservation call, 24/7, with a warm, on-brand greeting
- Take a complete booking (dates, guest count, room type, and property) and confirm it autonomously, with no human in the loop
- Absorb peak demand without adding front-desk headcount
- Get measurably better over time
Month one, in numbers
In the first full month live, the voice agent delivered:
| Outcome | Month one |
|---|---|
| Guest calls handled with no human on the line | 900+ |
| Confirmed reservations booked end-to-end by the AI | 52 |
| Booking revenue through the voice line | $55,000+ |
| New-reservation conversion (booking-intent calls only) | 16.3% |
| Calls connected and completed | 99.2% |
| Reservation-handling accuracy, week 4 | 66% |
| Unique guests reached | 600+ |
Demand on the line grew as guests started using it. Daily volume went from about 15 calls at the start of the month to about 46 by the end, a roughly threefold increase, with no extra people on the desk.
How 16.3% was counted
That conversion rate is not 16.3% of every call. A reservations line takes more than new bookings. Guests also rang about existing stays, changes, cancellations, and general questions. Those calls were never going to produce a new reservation, so they do not belong in the denominator.
We counted conversion only on booking-intent calls: the guest asked to make a new reservation. Of those callers, 16.3% left with a confirmed stay written by the agent. That is about one confirmed booking for every six guests who rang to book.
The 52 confirmed reservations are the ones that made it all the way through: property, dates, guests, room type, live availability, and confirmation. Callers who were still deciding, who declined alternative dates when the property was full, or who ended the call before a room was confirmed sit in the remainder, not in the 52.
What the accuracy number measures
99.2% is line reliability. The call connected and completed without dropping. Fewer than 1% of calls were lost.
66% is different. It is an internal QA score of reservation-handling: did the agent take the right property, dates, guest count, and room type against live availability, and confirm it correctly? Week one scored 58%. By week four it was 66%, an eight-point lift as the agent learned the group's inventory, sold-out patterns, and policies.
That is a month-one trajectory, not a finished ceiling. The useful signal is the slope: accuracy improved every week, on its own, with no extra headcount.
Why it converted
- A consistent, warm greeting and a natural, human-sounding conversation on every call
- The full booking flow handled start to finish: location, dates, guests, room selection, confirmation, and live availability
- Fully booked dates handled without a dead end. The agent offered alternative properties and dates instead of leaving the guest stuck
- Complex, multi-room group bookings managed without confusion
- A clear sense of scope. Requests about existing bookings, changes, and cancellations were recognised and routed correctly every time
The takeaway
In 31 days, 24/7 autonomous reservations showed up as confirmed bookings, real revenue, and a system that got sharper each week. No extra headcount. No closed window. No missed call.
See what an enterprise voice agent could do for your properties