Case study
AI customer service software for an e-commerce brand
An AI system that answers routine customer messages right away and hands complex cases to a person with the full conversation attached. Average response time dropped from 4 hours to 3 minutes, and customer service staffing costs fell 60%.
- An e-commerce brand with more than 200,000 members
- E-commerce

- Response time: from 4 hours to 3 minutes
- 60% lower customer service staffing cost
- Customer satisfaction up to 4.9 out of 5
- 85% of inquiries answered automatically, around the clock
What was getting in the way
- Hundreds of messages a day: even five agents could not keep up, and peak wait times went past four hours
- More than 70% of messages were repeat questions (order status, shipping, returns) that used up most of the team’s time
- Nobody answered in the evening or on weekends, so sales were lost
- Answers varied from agent to agent, with no shared standard or wording
- There was no data on common questions or customer pain points
How the system solves it
- A natural-language engine that recognizes and answers 12 categories of common questions, including orders, shipping and returns
- Smart handoff: anything the AI cannot handle goes to a person with a conversation summary, so nothing gets lost
- A knowledge base console where the support lead edits AI answer templates without an engineer
- The LINE official account, website chat and Facebook Messenger in one shared inbox
- A live dashboard tracking AI resolution rate, handoff rate, response time and satisfaction
What is inside
Natural-language understanding
Understands conversational Traditional Chinese and recognizes 12 intent categories such as order status, returns and product questions.
Smart human handoff
Hands the conversation to an agent when the AI cannot resolve it, with a summary and sentiment analysis attached.
Knowledge base console
The support lead maintains FAQs, answer templates and product information, and the AI picks up changes right away, with no engineer involved.
One inbox for every channel
LINE official account, website chat and Facebook Messenger, all managed in one place.
Live monitoring dashboard
Tracks AI resolution rate, average response time, handoff rate and customer satisfaction.
Sentiment detection and alerts
Spots negative sentiment as it happens, flags high-risk conversations and routes them to senior agents first.
What changed
- Average response time fell from 4 hours to 3 minutes, so customers no longer leave while they wait
- The team went from five agents to two, saving about NT$1.8 million a year in staffing costs
- The AI resolves 85% of inquiries on its own; only 15% need a person
- Customer satisfaction rose from 3.6 to 4.9 out of 5, and negative reviews fell 70%
- Conversion on after-hours orders rose 40% with answers available around the clock
How the project came together
The client is an e-commerce brand with annual revenue above NT$100 million and more than 200,000 members, and it receives hundreds of customer messages a day. Five agents answered them by hand through the brand’s LINE official account. At peak times customers waited four hours on average, which cost sales and brought in bad reviews.
We built an AI customer service system around a natural-language understanding engine. It recognizes common requests such as order status, returns and exchanges, and product questions, and answers them automatically. Anything the AI cannot resolve is handed to a human agent together with the full conversation, so the agent picks up exactly where the AI left off.
After launch, response times dropped sharply and the team went from five agents to two, who now focus on high-value customers and complex cases.
Other systems we have built
Want a system like this?
Tell us how your process works today. We will scope a version for your team, with a requirements document, a detailed quote and a timeline.