AI Zendesk Support with RAG, OpenAI & Supabase KB

This workflow triggers on new Zendesk tickets via webhook, extracts ticket details like subject, description, and requester info. It then queries a Supabase vector store with OpenAI embeddings to retrieve the most relevant knowledge base articles. An AI RAG agent drafts professional replies incorporating retrieved KB content and Postgres-stored conversation history.

Decision logic checks for relevant KB matches; if none or sensitive topics (e.g., KYC, refunds), it sends a fallback response, tags for human review, and logs. Otherwise, it posts the AI reply, tags with 'ai_reply', and updates memory for context in future interactions.

Benefits include 24/7 automated support, reduced response times, consistent professional replies, and escalation only when needed. Ideal for customer support teams handling high ticket volumes, minimizing manual effort while maintaining quality via internal KB.

Use cases: E-commerce helpdesks, SaaS support, internal IT/service desks scaling with AI without hiring more agents.