Hospitality · AI assistant

Villa Levante

A multilingual AI assistant that responds to guests 24/7 and turns enquiries into direct bookings.

Sector
Hospitality
Role
Design + full-stack development
Type
AI assistant · Web
Status
Demo project · live prototype
villalevante.it
The context

A property receiving guests from around the world, at every hour of the day.

Enquiries arrive in multiple languages and outside office hours; every late response is a booking that risks shifting to an OTA, along with its commission. Staff spend hours on repetitive questions — availability, services, directions — taking time away from real hospitality.

The goal: respond immediately, in any language, and bring the booking back to the direct channel.

The solution · two-sided lever

Cut costs and increase direct revenue at the same time.

Cost side

Fewer hours, no opening times

The assistant handles recurring questions autonomously, in any language, 24 hours a day. Staff step in only when it genuinely matters.

  • Instant responses day and night
  • Native multilingual, no translators needed
  • Hours freed from repetitive enquiries
Revenue side

More direct bookings

By responding promptly and well, the assistant guides the guest to book on the direct channel — reducing OTA dependency and their commissions.

  • Guided direct conversion
  • Potential OTA saving (15–25%)
  • Guest relationship from first contact
24/7
Request coverage, every time zone, every language.
up to 25%
OTA commission potentially recovered per direct booking.
hrs/week
Staff time returned to real hospitality.
Stack

Built to last and integrate.

A conversational assistant connected to real property data, on a lightweight and fast web infrastructure. No dependency on closed platforms: everything stays yours.

Conversational LLMRAG on property dataMultilingualNext.jsVercel AI SDKSupabase
Live demo

Try the working prototype.

This is a functional prototype — not a real hotel. It lets you see concretely how the assistant works, how it responds, how it guides the guest.

Next project

Softale

Softale