Four languages, one overwhelmed support team
A regional EU travel agency was handling 900+ customer queries every week across Dutch, French, English, and German. The four-person support team was averaging a 4-hour reply time and a 71% CSAT score. Most queries were repetitive: visa requirements, itinerary changes, baggage rules, cancellation policy.
Translation friction made it worse. A French-speaking customer asking about visa rules for Vietnam meant a support agent reading the English-language carrier documentation, drafting the answer in French, and praying their translation matched the legal text.
A multilingual RAG chatbot, sourced from the agency's own knowledge
We built a chatbot grounded in the agency's destination guides, fare rules, visa requirement archive, T&Cs, and historical FAQ tickets. Language is detected automatically from the customer's first message. No setting, no toggle.
When the chatbot cannot resolve a query (about 21% of cases), it hands off to a human agent in the customer's language with the full conversation context attached. The handoff feels seamless because it is.
Tickets down, satisfaction up
Auto-resolve climbed to 79%. Support ticket volume dropped by 64% in the first two months, which let the existing four-person team move from triage mode to proactive customer success work.
CSAT recovered to 88% within the same period. The improvement came from speed: a question answered in 4 seconds beats a perfect answer 4 hours later.