All Projects
AI Client Conversation Router
AI-Powered Client Conversation Router
An intelligent conversation triage system that receives client conversation transcripts via webhook, uses OpenAI GPT-4o-mini to generate concise summaries, then routes the conversation to the correct department (Support, Administration, or Commercial) via email. Simultaneously, the workflow posts a notification to Slack, sends a WhatsApp Business message to the client confirming receipt, searches HubSpot for the client's contact record, and saves the AI summary as an engagement note -- creating a complete audit trail across all channels.
Tech Stack
| Layer | Technology |
|---|---|
| Automation Platform | n8n |
| AI / LLM | OpenAI GPT-4o-mini (Summarization + Routing Agent) |
| CRM | HubSpot (Contact search + Engagement notes) |
| Email Routing | Gmail (department-specific routing) |
| Team Notifications | Slack (OAuth 2.0) |
| Client Notifications | WhatsApp Business Cloud API |
| Trigger | Webhook (POST /client-conversation) |
Workflow Architecture
Webhook (POST with transcript)
-> Define Routing Emails (Support / Admin / Commercial)
-> AI Summarization (GPT-4o-mini, 2-3 sentence summary)
-> [Parallel Branch 1] Slack Notification -> WhatsApp Confirmation
-> [Parallel Branch 2] HubSpot Search Client -> Save Notes as Engagement
-> [Parallel Branch 3] Router Agent (GPT-4o-mini) -> Gmail (route to department)Business Outcomes
| Metric | Impact |
|---|---|
| Conversation routing accuracy | AI-driven classification eliminates human misrouting; routes to Support, Admin, or Commercial based on content analysis |
| Response time to client | Instant WhatsApp confirmation vs. hours of manual acknowledgment |
| CRM data completeness | 100% of conversations automatically logged in HubSpot with AI summaries |
| Department email accuracy | Zero misrouted emails -- AI reads transcript context, not subject line guesses |
| Multi-channel coverage | Single conversation triggers updates across Slack, WhatsApp, HubSpot, and Gmail simultaneously |
| Manager oversight | Slack channel provides real-time feed of all incoming client conversations |
Technical Metrics
| Metric | Value |
|---|---|
| Nodes in workflow | 11 |
| AI model calls per execution | 2 (summarization + routing agent) |
| AI temperature | 0.2 (summarization), 0.1 (routing) -- low for consistency |
| External integrations | 5 (OpenAI, HubSpot, Gmail, Slack, WhatsApp) |
| Parallel branches | 3 (notification, CRM logging, department routing) |
| Language support | Multi-language (responds in same language as conversation) |
Estimated Cost Savings (If Implemented)
| Item | Manual Cost | Automated Cost |
|---|---|---|
| Receptionist/dispatcher time per conversation (10 min @ $20/hr) | $3.33/conversation | ~$0.01 (OpenAI API) |
| 200 conversations/month | $666/month | ~$2/month |
| CRM logging time (5 min/entry) | $1.67/entry | $0 |
| Misrouted conversation cost (rework) | ~$15/incident | $0 |
| Annual savings (200 conv/mo) | -- | ~$8,000-$10,000/year |