The Problem
Lohono Stays manages ultra-premium villa rentals across India, with nightly prices ranging from Rs 60,000 to Rs 2,50,000. Many callers were high-intent travelers reaching out from different time zones, often well outside normal business hours.
But the operation had a critical gap: no representatives were available overnight. Calls went unanswered, enquiries dropped, and every missed ring created real revenue risk. At these booking values, a single lost conversation could mean a Rs 10-15 lakh opportunity disappearing before the team even woke up.
The Old Way
Guests called after hours, heard silence or voicemail, and had no reliable way to continue the booking journey until the next day.
The AI Way
The system answered instantly, qualified intent, shared relevant stay details, captured contact information, and locked in a morning callback window every time.
The Product Experience
The design goal was not just automation. It was reassurance. A guest calling about a premium stay at 2 AM should still feel like they had reached a capable hospitality brand, not a dead end.
Caller Journey
The experience felt like one natural conversation, but behind it was a carefully designed orchestration layer routing different intents to specialized flows without exposing that complexity to the caller.
System Architecture
The system was structured around a simple operational promise: every after-hours call should end with useful progress and a clear next step. That required routing logic, specialist handling, and graceful fallbacks rather than a single generic voice bot.
Triage and router for every incoming call.
Identified whether the caller needed availability help, property-specific information, booking support, or non-business handling.
Availability specialist for location-first enquiries.
Collected destination, travel dates, group size, and budget when guests asked broad questions like "Do you have anything in Alibaug?"
Property-details specialist.
Handled property-specific pricing, confirmed stay details, and captured contact information for clean follow-up.
Booking support flow.
Served callers with existing reservation-related questions so the experience remained useful for current guests too.
Non-business handler.
Gracefully absorbed off-topic or unproductive conversations without breaking the overall caller experience.
Build Snapshot
The build ran from August to October 2025 and was deployed as a live pilot rather than a lab demo. That meant the product had to work across real noise, unclear intent, and real booking pressure from day one.
| Dimension | Detail |
|---|---|
| Build period | August - October 2025 |
| Pilot duration | 7 weeks live |
| Agents deployed | 5-agent orchestration |
| Call volume | Approximately 300 calls per day |
| Accuracy | 95% real-world accuracy |
| Scenario coverage | 90% handled end-to-end |
| Knowledge coverage | 93% at launch, improved weekly |
| Property range | Luxury villas priced from Rs 60,000 to Rs 2,50,000 per night |
Pilot Results
The pilot validated that the product could handle real hospitality demand, not just ideal scripted paths. Accuracy reached 95 percent across all calls. The remaining edge cases were mostly callers in highly noisy environments, and even then the system responded with a callback promise rather than a dead end.
Ninety percent of scenarios were completed end-to-end by the AI. The remaining 10 percent were escalated gracefully, with caller details preserved and a next-day follow-up window confirmed so no lead was dropped.
One product insight shaped the system during the pilot: most guests started location-first rather than property-first. Updating the routing logic to prioritize destination discovery reduced unnecessary back-and-forth early in the conversation and made the experience feel smarter immediately.
"Do you have anything in Alibaug?" turned out to be the dominant pattern, not property-specific search.
The routing logic was updated during the pilot to prioritize location-first discovery before property qualification.
Client Outcome
The AI Became the Night Shift
The product manager reviewed the pilot results and approved full handover. No missed calls. No lost leads. No human required between 8 PM and 10 AM.
That outcome mattered because the system was protecting high-value booking intent, not low-value support volume. At an average booking value around Rs 1.5 lakh, recovering even a handful of enquiries per week represented meaningful revenue protection.
Product Skills Demonstrated
Multi-Agent Orchestration
Designed five specialized agents to behave like one coherent hospitality experience.
Conversation Architecture
Built branching voice flows with graceful fallbacks and reliable next-step capture.
High-Stakes Voice AI
Shipped for premium clientele, international callers, and revenue-sensitive booking moments.
Knowledge Operations
Tracked answer gaps weekly and improved coverage continuously during the live pilot.
Pilot to Production
Moved from live pilot to full client handover with operational approval.