Telegram Lead Capture & Postgres CRM with n8n & Groq

Turn Telegram messages into a structured CRM lead—instantly

This n8n workflow captures leads from Telegram (voice notes, typed lines, or business-card photos), uses Groq AI to extract structured details, checks Postgres for duplicates, and sends one-tap confirmation buttons back in Telegram. A daily 8:30 digest then surfaces who needs follow-up.

What this workflow does

• Receives lead input from incoming Telegram messages or button taps, and also runs on a daily 08:30 cron to generate follow-up digests and clean up expired pending confirmations.
• Validates the sender against an allowed chat ID list; unauthorized senders receive a reply containing the chat ID.
• Converts the input into text:
  • Transcribes Telegram voice notes with Groq Whisper
  • Reads business-card images with Groq Vision
  • Uses typed messages as-is
• Extracts lead fields using Groq Chat Completions, including contact, company, need, budget, urgency, follow-up date, and stage update—then validates/normalizes the result.
• Prevents duplicates in Postgres by searching for up to three likely matching leads and sending a Telegram confirmation message with inline buttons to:
  • update an existing lead
  • save a new lead
  • cancel
• On button tap, claims the pending note from Postgres, creates/updates the lead, and replies with saved lead status, stage, score, and follow-up date.
• Runs /today and the scheduled digest to query Postgres for due/overdue follow-ups and cold leads, then sends a formatted summary to Telegram.

Use cases

• Capture sales/partnership leads from real conversations you record as voice notes.
• Batch-process networking contacts from business-card photos into a Postgres-backed CRM.
• Daily follow-up reminders delivered automatically via Telegram.

Technical details

• Integrations/Nodes: Telegram, Postgres, Groq (Whisper, Vision, Chat Completions), plus n8n logic nodes such as if, switch, code, no op, and postgres.
• Core workflow: normalize Telegram updates → AI extract lead data → Postgres duplicate match (top 3) → inline confirmation → CRM write-back.