>95% answer accuracy across five-language guest service
A world-renowned integrated resort in Singapore deployed a multilingual AI Knowledge Assistant from Dyna.Ai for a customer service operation handling guest enquiries across phone and email. The solution helps agents retrieve relevant information and prepare consistent, ready-to-use responses across five languages.
The team handles enquiries across hotel operations, housekeeping, restaurant reservations, room information, retail shops, and other resort services. The business problem was not a lack of information. It was finding the right information quickly while a guest was waiting, particularly when the answer could require several operational sources.
Enquiries arrive in English, Chinese, Japanese, Korean, and Thai. Agents needed one consistent knowledge foundation across languages and channels so the answer would not depend on which source an agent searched or how the agent translated the response.
The resort needed an internal assistant that could retrieve relevant information, generate a ready-to-use response with reference materials, and help agents maintain consistent service without removing them from the guest relationship.
The challenge
Guest enquiries spanned five languages and many resort services
Dyna.Ai implemented a multilingual, retrieval-augmented AI Knowledge Assistant for the internal customer service workflow. When a guest question arrives, the assistant retrieves relevant content from a centralized knowledge base, prepares a response, and presents the supporting references for agent review.
The assistant supports English, Chinese, Japanese, Korean, and Thai across phone and email service. Agents can review and adapt the response for the guest, while escalating when the request requires action or judgment from another resort team.
The assistant was not positioned as a generic answer generator. It was a retrieval and response-preparation layer grounded in the resort’s knowledge base. The agent remained responsible for what was communicated and for the next service action.
The deployment
One knowledge layer across phone and email service
The assistant was introduced into the existing customer service operation rather than as a separate guest-facing channel. The resort remained the owner of the guest experience, while service agents used the assistant to bring relevant knowledge closer to live interactions.
The workflow covers guest enquiries across phone and email, with multilingual support for five languages and a broad resort knowledge base spanning accommodation, dining, housekeeping, retail, and other services.
Implementation depends on knowledge ownership. Room information, amenities, booking rules, service hours, and escalation instructions may change. The assistant therefore brings source materials into the response workflow so agents can verify information and open the original content when needed.
How the workflow changed
Language-agnostic retrieval kept one workflow across five languages
User queries are normalized into English before the knowledge base is searched. This creates a common retrieval path regardless of the language used by the agent, while the final response is generated and translated back into the user’s original language.
Images, PDFs, URLs, and other source materials can be included as accessible references in the generated response. Agents can move from a guest question to a source-backed answer more directly, verify the information, and decide when another team needs to take action.
The Knowledge Assistant gave the service team one multilingual path from question to response. Agents could retrieve relevant information without manually searching across multiple sources, use a ready-to-review answer, and retain the source context behind the response.
The results
Ready-to-use, source-linked responses improved consistency
The resort gained a shared knowledge and response workflow across agents and languages. Ready-to-use responses reduced the effort required to research and formulate answers, while direct links to images, PDFs, URLs, and other source materials strengthened traceability.
The reported result is more than 95% answer accuracy across the multilingual guest-service operation. Agents can work from the same knowledge foundation, respond in the guest’s language, and verify the source before sending or adapting the answer.
