Dmytro Khomenko

AI-Powered Email Automation (n8n, GPT Agents, Pinecone, Supabase)

Self Employed • Freelance

Jul 2026 - Jul 2026 • 0 mos

An online handmade cosmetics store needed full automation of inbound email processing to reduce team workload (1–3 people, 15–30 emails/day), speed up response times, and prevent important messages (orders, invoices, support/sales inquiries) from being missed. What Was Built AI Email Classification & Routing Built an n8n workflow with an AI Text Classifier that analyzes every Gmail message and sorts it into 5 categories: Internal, Customer Support, Promotions, Admin/Finance, Sales Opportunities. Each email is auto-labeled and routed to the matching AI agent. Brand Knowledge Base (RAG) Created a vector knowledge base in Pinecone from the brand's product catalog, pricing, shipping terms, and FAQs. AI agents query it before drafting replies, ensuring accurate, brand-specific responses. Automated Reply Generation Each category has a dedicated AI agent drafting replies directly in Gmail. Staff just review and hit "Send" — no manual writing needed. Conversation Memory Wholesale/commercial inquiries are logged in Supabase (Postgres) with contact info and message history, letting the Sales agent retain context across conversations. Logging & Reporting Every email is logged in Google Sheets and Supabase (date, sender, category, subject, action), giving the owner full visibility with zero manual tracking. Real-Time Notifications: A Telegram bot instantly alerts the owner about emails needing human attention (complaints, wholesale inquiries, finance issues). Results Email is now processed automatically, 24/7: every message is classified, labeled, and drafted with accurate brand data where needed. Leads are captured for follow-up, urgent messages are never missed, and inbound email stats are tracked automatically — with zero manual effort from staff.