AI Travel Concierge — Lotus Hotels & Resorts
Client
Lotus Hotels & Resorts, a hospitality group with properties across Mumbai and the Konkan coast.
Problem
The hotel website answered the same questions repeatedly — check-in times, parking, pool hours, airport transfers, local sightseeing — and guests with urgent queries phoned the front desk, tying up staff and overflowing the call queue during peak season. Direct website bookings were being lost to OTA channels because the site could not answer even simple questions in real time.
Solution
Gravitas India deployed an AI travel concierge on the hotel website: a large-language-model chatbot grounded in the hotel's own content — property pages, rate cards, FAQs, and local-area guides — so answers were accurate and on-brand rather than generic. The concierge answered instantly, escalated to a human only when a query needed a live agent, and captured booking intent by handing off to the reservations flow at the right moment.
Implementation / Technical Approach
- RAG-grounded responses — retrieval-augmented generation over the hotel's content corpus, so every answer cited a source page and hallucination risk stayed near zero.
- Booking-intent detection — the model classified each conversation as informational, booking-ready, or service-request, routing booking-ready guests straight to the reservations widget.
- Human hand-off queue — service requests (room service, late check-out, maintenance) were triaged into a front-desk queue with full conversation context, so staff never re-asked.
- Multilingual support — English, Hindi, and Marathi in a single deployment, matching the property's guest mix.
- Usage analytics — every unanswered question was logged and fed back into the content corpus, so the concierge got measurably smarter each month.
Technology
LLM RAG Vector Search Intent Classification Web Widget
Outcomes
Routine enquiries stopped reaching the front desk — 87% of conversations were resolved autonomously, and call volume fell 62% within the first quarter. With booking-ready guests handed straight to the reservations flow, direct website bookings rose 31% against the same period the year before, reducing OTA commission spend. The hand-off queue kept service requests fast, and monthly feedback from unanswered questions meant the concierge's answer rate improved with every content refresh.