Customer-Service AI for International Guests

Within the accommodations business of a multi-business group, we delivered a "customer-service AI" to address inbound demand. We trained the AI on each property's ambience, service offerings, and prior interaction history to generate individualized AI clones for each location. We built a system in which the AI handled first-line responses to multilingual inquiries — in English, Chinese, and other languages. This reduced the time front-desk staff spent on phone and email inquiries and created an environment in which they could focus on on-site hospitality. We also implemented a Human-in-the-Loop (HITL) process in which staff refined AI responses, enabling the AI to continuously learn and improve.

[Challenges]

• Multiple resort properties were generating high volumes of inquiries (driven by inbound demand, largely in Chinese and English).
• Each property's assigned staff had to handle inquiries case by case (know-how was concentrated in those staff members, creating dependence on individual experience).
• Staff could not focus on their core operational work (because they had to constantly handle inquiries across multiple languages).

[Results]

• Generated individualized AIs for each resort property and trained them on the interaction record (the AIs learned the response history across multiple languages so it could be reproduced).
• Each property's individualized AI absorbed inquiries and grew over time (handling not only basic items such as directions but also responses on ambience and service).
• The AI provided answers and humans refined the nuances. HITL and workload reduction were achieved together (refinements became ground-truth data, so the AI grew smarter with use. Staff workload was reduced by 30%).