PRODUCT
AI Customer Service
More than a scripted chatbot: it uses the enterprise knowledge base as its source of truth, understands colloquial questions and multi-turn context, and can look up orders, reschedule bookings and create tickets in conversation. It assesses confidence and risk before handing over to a human with a summary, and unresolved questions become new knowledge.
Try it onlineAI customer service capabilities
With the enterprise knowledge base as the source of truth, answers retrieve before generating, so policy, price and process changes take effect the same day.
Knowledge baseSourced answersUnderstands colloquial questions, abbreviations and multi-turn follow-ups, so users never need to start over when they change their mind or add details.
Language understandingMulti-turn dialogueConfidence and risk decide between self-service and hand-off, with context and verified details passed to the agent.
Human-AI collaborationSmart handoffProactively reaches out based on order, ticketing, weather and inventory events, flagging reschedules, refund progress and travel changes before users ask.
Proactive serviceEvent triggers
How it differs from standard support
Standard support is a script library with human fallback; AI customer service understands, retrieves, acts and keeps learning
| Dimension | Standard support | AI customer service |
|---|---|---|
| Understanding | Keyword matching with fixed scripts fails as soon as a question is phrased differently | AILLMs understand intent and context — colloquial wording, typos and multi-turn follow-ups included |
| Knowledge source | Scripts configured line by line by hand, so policy and price changes often lag behind | AIDocuments, FAQs, product and order fields are chunked and indexed automatically; updates take effect immediately with sources cited |
| Problem solving | Only answers questions; rescheduling and refunds are handed off elsewhere | AILooks up orders, reschedules, issues invoices and creates tickets in the conversation — answering and acting in one flow |
| Complex cases | Hands off at the first edge case, and users must explain everything again | AIRoutes by confidence and risk, giving agents a summary, verified details and suggested replies |
| Service hours | Depends on staffing schedules; queues at night and during peaks | AIOnline 24/7, scaling automatically at peak and covered by AI at night |
| Operations | Manual spot checks of chat logs; issue types inferred from experience | AIFull-volume QA, attributing intents, sentiment and unresolved hotspots to knowledge entries automatically |
It doesn't just chat — it gets things done
Connecting knowledge, business data and human agents in one service flow
Unified omnichannel access
Mini programs, official accounts, H5, websites, WeCom and Douyin DMs share one conversation and agent workspace, so context follows users across channels.
Knowledge base & retrieval
Enterprise documents, FAQs, ticketing policies and product information are chunked and indexed automatically; retrieval precedes generation, with sources attached to reduce hallucination.
Intent recognition & routing
Pre-sales enquiries, changes and refunds, and complaint risk are classified automatically and routed to the right skill group and ticket queue.
Business actions
Check orders and availability, request a reschedule, issue e-invoices and track refunds inside the conversation, calling APIs to complete them when needed.
Human-AI collaboration
AI responds first and agents can take over at any time, receiving the conversation summary, verified details and suggested replies — no need for users to repeat themselves.
Quality assurance & risk alerts
Real-time alerts on commitment language, sensitive words and rising sentiment; full-volume QA with disputes traceable to the original conversation and knowledge sources.
EXAMPLE
One conversation, one service completed
Users never have to ask for a human or repeat their order number. AI identifies intent, checks knowledge and orders, completes what it can, and hands the rest to an agent with full context.
- Detects intent and sentiment, flagging complaint risk
- Retrieves knowledge before answering, citing sources for key conclusions
- Completes order lookups, reschedules and other actions in the conversation
- Hands low-confidence or high-risk cases to agents with a summary
Service loop
From a single question to knowledge updates, every conversation makes the next one more accurate
- 01
User asks
From a mini program, official account or website — text, images and order numbers all accepted.
- 02
Intent detection
Determines whether it's an enquiry, a transaction or a complaint, reads sentiment and urgency, and decides what happens next.
- 03
Knowledge retrieval
Retrieves policies, products and order data by intent, establishing the basis before wording the answer.
- 04
Act or hand off
Self-service actions complete directly; out-of-scope or high-risk cases go to an agent with a summary.
- 05
Feedback loop
Unresolved and low-confidence questions feed back as knowledge gaps, so the base gets more accurate with use.
Use cases
High-volume, repetitive, rule-based enquiries are the first to be handled by AI
Attractions & parks
Ticket rules, rescheduling and refunds, entry notices and availability checks — AI absorbs the enquiry peak around holidays.
Hotels & guesthouses
Instant answers on availability, check-in times, invoices and deposits, with proactive alerts and guidance on order exceptions.
Ecommerce & retail
Shipping status, return policies and coupon usage, with after-sales tickets created automatically and results written back.
Distributors & channels
Commission settlement, asset access and ordering processes — recurring channel questions handled by the knowledge base and AI.
