MR. LIVING Uses Crescendo Lab CAAC AI Agent to Answer and Route LINE Customer Messages

MR. LIVING, a Taiwanese furniture brand known for Nordic-inspired designs, deployed Crescendo Lab's CAAC conversational AI platform on its LINE Official Account, which receives more than 16,000 messages a month. The AI Agent now answers 75% of routine messages directly, routes sensitive messages to human agents with 81% accuracy, and was estimated during testing to save the customer service team 160 hours of repetitive work each month. "We don't want our customer service agents to be trapped in a loop of answering FAQs and have no time to understand what customers really need," said CTO Lin Yu-Fan. Before the project, the team analyzed more than 10,000 LINE conversations. 86.8% of customers' first messages were text, and 43% were basic questions about store information, invoices, payments, warranties and deliveries, which took up nearly half of agents' working hours. The other 57% were more complex needs such as order changes, repairs, promotions and styling advice that span customer service, sales and design consultants, so customers often had to repeat themselves as messages were passed between teams. MR. LIVING set the rule that AI handles routing and information structuring while humans handle judgment and reassurance. The agent classifies each incoming message and answers common questions such as store hours, delivery status and warranty terms. Topic control and keyword detection hand emotional or sensitive conversations, such as returns, delays and repairs, straight to a person. Crescendo Lab replaced the brand's CSV-based FAQ process with a Chatbase-integrated backend where staff upload links and files and flag wrong answers for correction. A cross-department "AI Teacher Team" maintains the agent: engineers mine conversation data for frequent issues, customer service managers adjust knowledge logic and categories, and marketing and design staff test tone. The brand is starting with a co-pilot model in which AI assists rather than replaces agents, and is gradually extending it to store sales teams. It plans to add an AI Sales Agent that recommends products and pushes product cards during conversations once the support agent's performance is stable.