Qualify email leads and create CRM deals
An n8n agentic workflow classifies incoming emails, qualifies leads, creates CRM deals and handles follow-ups, storing conversations in Supabase. ·
Done withn8n
- What they did
- They built a self-hosted n8n workflow around a shared Gmail inbox for a procurement client. Incoming emails are classified with OpenAI, key signals like budget, timeline, project size, authority, and readiness are extracted, and the lead state is stored in Supabase so later emails update the same record; deterministic rules then route follow-ups, escalate when needed, and create CRM customer and deal records through the CRM API once enough information is gathered.
- How it went
- For a business handling about 200 inquiries a month, the first month produced roughly 3% qualified leads, and the author also said around 3% of qualified inquiries converted into closed business. It still uses human review for quote outputs on hot leads and for escalations or edge cases.
- Worth knowing
- The author says the hardest part was getting the persistent memory architecture and lead-scoring logic reliable, and that the workflow shown was only the orchestration layer over multiple sub-workflows.
Try it yourself with n8n
Build an n8n agentic workflow that watches [inbox] for new emails, classifies each as lead, support, or other, and qualifies leads using [criteria, e.g. budget, company size, need]. For qualified leads, create a deal in [CRM] and draft a follow-up reply, storing the conversation history in [database like Supabase]. Save follow-ups as drafts for my approval instead of sending them.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: Reddit · Jun 10, 2026
Related use cases
Add todo items to Todoist from Telegram
A Telegram bot with an LLM and an orchestration platform (n8n or windmill) adds todos to the author's Todoist account from natural language messages.
Catalog and summarize papers dropped into Drive
An n8n workflow extracts metadata from new PDFs, finds related papers on Semantic Scholar and PubMed, builds an APA citation, logs to Sheets and creates a Notion summary.
Qualify leads and draft emails pending human approval
An n8n workflow dedupes and AI-scores leads and drafts personalized emails, sending via Gmail only after human approval.
Sell services through an agent that takes payment and delivers
An agent runs a transaction flow, creates Stripe checkout links, verifies payment before delivering a service, and asks the owner on Telegram to approve refunds or disputes.
Custom agent
Answer home-service customer calls and text the owner
A voice agent answers calls for home-service businesses, records caller details and request, and texts the owner a status.
Custom agent
Enrich new CRM leads for personalized outreach
A Make.com automation enriches new leads entering the CRM with detailed profiles so cold outreach can be more targeted.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.