AI 客服助理儀表板AI customer-service assistant dashboard
為虛構產品「訊衡 Desk」設計的 AI 客服儀表板概念：AI 分類 LINE 等管道的訊息並草擬回覆，由真人核准後才送出，再以分析頁找出常見問題。
An AI service desk concept for Xunheng Desk, a fictional product: AI sorts messages from LINE and other channels and drafts replies, a person approves them before they are sent, and an insights page surfaces recurring questions.
此為概念設計，用於展示我們的設計與開發能力，並非真實客戶案例。
This is a concept design that demonstrates our design and development capabilities; it is not a real client project.
中小品牌的客服訊息分散在 LINE 官方帳號、網站表單與 Email，同樣的問題每天重複回答；直接交給 AI 自動回覆，又擔心答錯或語氣不對。
這個概念把 AI 定位成助理而不是代理：收件匣依 AI 分類與信心分數排序，每則草稿都標示引用的訂單、物流或知識庫來源，客服可以一鍵核准、修改或轉給同事。涉及退款、個資與客訴的訊息，一律由真人核准後才送出。
技術上會以 Next.js 建置後台，串接 LINE Messaging API 與網站表單，使用大型語言模型 API 分類與草擬回覆，並以檢索增強生成（RAG）讓回覆只引用品牌自己的常見問題與政策文件。分析頁中的數字為示意資料。
Smaller brands get customer messages through a LINE Official Account, website forms and email, and answer the same questions every day. Handing everything to an AI auto-responder raises a different worry: wrong answers, or the wrong tone.
This concept treats the AI as an assistant, not an agent. The inbox is sorted by AI category and confidence, every draft shows the order, shipping or knowledge-base source it relies on, and staff can approve, edit or reassign it in one click. Anything involving refunds, personal data or complaints always needs a person to approve it.
Technically it would be a Next.js back office connected to the LINE Messaging API and website forms, using a large language model API to classify and draft, with retrieval-augmented generation (RAG) so replies only cite the brand’s own FAQs and policies. Figures on the insights page are illustrative.
