PromptForge
Back to list
AI应用WhatsApp客服机器人对话设计意图识别多轮对话

WhatsApp AI 客服机器人多轮对话与意图路由设计模板

设计WhatsApp AI客服机器人的完整对话流程,包括意图识别、多轮对话状态管理、人机协作升级策略和多语言支持方案。

6 views4/28/2026

You are a Conversational AI Architect specializing in WhatsApp Business API chatbot design.

Design a complete WhatsApp AI customer service bot for my business.

Business Context

  • Business type: [e.g., e-commerce / SaaS / restaurant / clinic]
  • Daily message volume: [estimated]
  • Languages needed: [e.g., Chinese, English, Spanish]
  • Current support channels: [e.g., email only, phone + email]

Please Design

1. Intent Recognition and Routing

  • Top 15 customer intents with example utterances (3+ per intent)
  • Intent classification confidence thresholds
  • Fallback strategy when confidence is below threshold
  • Human handoff trigger conditions

2. Multi-Turn Dialogue Flows

For each top intent, provide:

  • Complete dialogue tree (user message -> bot response -> follow-up)
  • Slot-filling strategy for required information
  • Context carryover between turns
  • Edge case handling (user changes topic mid-flow)

3. Message Templates

  • Welcome message (first-time vs returning user)
  • Quick reply button layouts (max 3 buttons per message)
  • List message templates for product/service browsing
  • Rich media usage (images, documents, location)
  • Session expiry re-engagement message (24h window)

4. State Management

  • Conversation state machine diagram (Mermaid format)
  • Session storage schema (Redis/DB)
  • Timeout and cleanup policies
  • Multi-device session sync strategy

5. Human-AI Collaboration

  • Escalation criteria (sentiment, complexity, VIP)
  • Agent takeover UX (seamless transition)
  • AI-assisted agent responses (suggestions)
  • Post-handoff summary generation

6. Analytics and Optimization

  • Key metrics to track (resolution rate, CSAT, handoff rate)
  • A/B testing framework for response variants
  • Conversation log analysis pipeline
  • Monthly optimization review template

Provide actual message text examples in the specified languages, with WhatsApp-compatible formatting.