用 GPT-4 自动筛选并分配牙科线索到 GoHighLevel

将网站表单、Twilio 短信、Meta 广告线索统一汇入，经 AI 评分分类后自动同步 CRM、发送跟进消息并生成周报

给谁

需要自动化本地服务（牙科/医美）线索跟进的小型诊所或代理机构

你会得到

线索自动分类评分、高优先级即时响应、未响应线索定时再激活、每周数据报告

你需要

n8n · twilio · meta-ads · pinecone · gohighlevel · slack · google-calendar · postgres

步骤

1. 配置线索来源

   在 n8n 中配置三种触发器：复制 Webhook 节点 URL 到网站表单或落地页；连接 Twilio 触发器接收短信回复；连接 Meta Lead Ads 触发器并选择 Facebook 主页和线索表单。所有来源的线索将汇入 Merge Leads from All Sources 节点。

2. 连接服务凭证与默认值

   在 Set Workflow Defaults 节点填入诊所名称、GoHighLevel API 地址/令牌/Location ID、Twilio 发件号码、Slack 频道、Google Calendar ID、Pinecone 索引名、报告邮箱、高优先级分数阈值（默认 70）和跟进天数（默认 3）。在 n8n 凭证管理器中添加 OpenAI、Twilio、SMTP、Slack、Google Calendar、Postgres、Facebook、Pinecone 的真实凭证。

3. AI 线索评分与分类

   Prepare Lead Data 节点将三种来源的线索标准化为统一格式。Lead Qualifier Agent 使用 GPT-4 模型，结合 Calculate Lead Score 工具（确定性评分）、Check Calendar Openings 工具、Sync with GoHighLevel 工具和 Access Knowledge Base（Pinecone RAG）工具，输出结构化结果：治疗类别、0-100 评分、摘要、建议下一步。

   运行提示词 · 来源原文。Lead Qualifier Agent 节点的 text 字段（用户提示词）

   复制提示词

   =A new lead has come in for {{ $json.practice_name }}. Qualify this lead and recommend the right treatment pathway.
   
   Lead source: {{ $json.source }}
   Name: {{ $json.lead_name }}
   Phone: {{ $json.lead_phone }}
   Email: {{ $json.lead_email }}
   Stated interest: {{ $json.treatment_interest }}
   Message/notes: {{ $json.lead_message }}

   运行提示词 · 来源原文。Lead Qualifier Agent 节点的 options.systemMessage 字段（系统提示词）

   复制提示词

   You are a lead qualification assistant for a local dental/aesthetic practice. Your job is to read an inbound lead's details and:
   
   1. Classify the lead into exactly one treatment category: Implants, All-on-X, Cosmetic, Orthodontics, or General (use General if nothing else clearly fits).
   2. Score the lead 0-100 on likelihood to book a consultation, based on urgency signals, budget signals, and clarity of interest in the message. Use the Calculate Lead Score tool to compute this score consistently rather than guessing.
   3. Check calendar availability with the Check Calendar Openings tool if the lead expresses interest in booking.
   4. Look up any existing contact history in GoHighLevel with the Sync with GoHighLevel tool before treating this as a brand-new lead.
   5. Use the Access Knowledge Base tool to answer any specific treatment questions the lead raised, grounding your summary in real practice information rather than guessing.
   
   Always return your final answer in the structured format required. Be concise and factual — do not make medical claims or promises about outcomes, pricing, or timelines beyond what the knowledge base confirms.

4. 按治疗类别路由并同步 GHL

   Route by Lead Treatment 根据 AI 输出的 treatment_category 将线索分到 Implants、All-on-X、Cosmetic、Orthodontics、General 或 fallback 分支。随后通过 HTTP 请求查询 GoHighLevel：按手机号查找联系人、获取 pipeline stage、读取自定义字段，为后续动作提供完整客户上下文。

5. 触发跟进动作

   所有线索收到 Twilio 短信确认；有邮箱的线索收到 SMTP 邮件。Check Lead Priority 比较 lead_score 与 HIGH_PRIORITY_SCORE_THRESHOLD，高分线索触发：Twilio 语音通话（当前为静态 TwiML，可替换为 AI 语音）、Slack 高优先级告警、Google Calendar 预约占位。同时 Log to Database 将线索写入 Postgres，Facebook Graph API 回复 Meta 来源线索。

   运行提示词 · 来源原文。Post Alert to Slack 节点的 text 字段

   复制提示词

   =🔥 High-priority lead: {{ $('Prepare Lead Data').first().json.lead_name }} ({{ $('Prepare Lead Data').first().json.lead_phone }}) — {{ $('Lead Qualifier Agent').first().json.output?.treatment_category }}, score {{ $('Lead Qualifier Agent').first().json.output?.lead_score }}/100. {{ $('Lead Qualifier Agent').first().json.output?.summary }}

6. 定时再激活未响应线索

   Schedule Trigger 每 6 小时运行一次，从 GHL 拉取标记为 new-lead-no-response 且超过 FOLLOW_UP_DAYS_UNRESPONSIVE 天的联系人。Follow-Up Agent 使用 GPT-4 决定再跟进类型：Re-engage（直接短信/电话）、Nurture（温和邮件）、Voice Call（高价值回拨）或 Default。结果经 Parse Follow-Up Results 结构化后，由 Route by Follow-Up Type 分发到对应动作。

   运行提示词 · 来源原文。Follow-Up Agent 节点的 text 字段（用户提示词）

   复制提示词

   =This lead has not responded since initial contact. Decide the best follow-up action.
   
   Contact: {{ $json.contact?.firstName || $json.firstName }} {{ $json.contact?.lastName || $json.lastName }}
   Phone: {{ $json.contact?.phone || $json.phone }}
   Days since last contact: {{ $json.daysSinceLastContact || 'unknown' }}
   Previous treatment interest: {{ $json.customField?.treatment_interest || 'unknown' }}

   运行提示词 · 来源原文。Follow-Up Agent 节点的 options.systemMessage 字段（系统提示词）

   复制提示词

   You are a follow-up strategist for a local dental/aesthetic practice. For each unresponsive lead, decide exactly one follow-up type:
   
   - Re-engage: the lead showed real interest but went quiet — worth a direct, personalized SMS/call attempt
   - Nurture: the lead is early-stage or low-urgency — worth a gentle informational email instead of a hard sell
   - Voice Call: the lead is high-value or previously expressed urgency — worth a live callback attempt
   - Default: use this only if none of the above clearly fits
   
   Base your decision on how much time has passed, the treatment interest, and any conversation history in memory. Do not fabricate details not present in the record.

7. 生成并发送周报

   每周定时从 Postgres leads 表拉取最近 7 天记录，Aggregate Metrics for Dashboard 计算总线索数、高优先级数、平均评分、分类分布。Build Dashboard Report 格式化后，通过 HTTP 写入 GHL notes，并通过 SMTP 发送给团队邮箱。

你是一个 n8n 工作流搭建助手。用户想要复现一个牙科/本地服务线索自动筛选与分配系统。请先确认用户已准备好：n8n 实例、OpenAI API 密钥、GoHighLevel API 令牌、Twilio/SMTP/Slack/Google Calendar/Postgres/Pinecone/Meta 广告凭证。然后引导用户按步骤导入模板、替换 YOUR_* 占位符、搭建两个子工作流（日历查询、GHL 联系人同步），最后测试单个线索的完整流转。不要替用户执行任何可能产生费用的 API 调用。