Open-Source AI Lead Processing System: Building an Automated Sales Lead Pipeline with n8n + Ollama

This article introduces an automated sales lead processing system based on an open-source tech stack. Core components include the n8n workflow engine, local Ollama large model, Google Sheets lightweight database, and Telegram real-time notifications. The system enables unified access to multi-channel leads, intelligent classification and grading, automatic data entry, and real-time team push notifications. It aims to solve the pain points of low efficiency and easy omission in traditional manual lead processing, providing small and medium-sized enterprises with a zero-subscription-cost automation solution.

Traditional lead processing relies on manual screening and entry, which is inefficient and prone to missing key information. To address this issue, the project uses open-source technologies to build an automated pipeline, linking data standardization, AI analysis, persistence, and notifications to significantly improve the response speed of sales teams.

The system adopts a modular design and supports dual-channel input: Webhook (external channels) and Chat Trigger (built-in chat). The workflow includes: data standardization (unified into a normalized_payload object) → Ollama local model entity extraction (low temperature parameters of 0.1-0.2 to reduce hallucinations) → response parsing and fault tolerance (regex cleaning of JSON, filling missing fields with "Not specified") → Google Sheets data persistence and Telegram real-time notifications.

The system performs six-dimensional analysis of leads via the Ollama model:

1. Lead summary (Ukrainian-language request overview)
2. Lead temperature (Hot/Warm/Cold purchase intent level)
3. Market type (B2B/B2C)
4. Company information (extract organization name)
5. Contact information (identify phone number)
6. Budget range (parse financial boundaries) Structured extraction understands contextual semantics, making it smarter than keyword matching.

Deployment requires preparation:

• Infrastructure: n8n (Docker/Cloud), Ollama (same server or accessible), recommended models (llama3.2/gemma2)
• External services: Telegram bot (Bot Token + Chat ID), Google Sheets (OAuth2/service account)
• Configuration steps: Import n8n workflow JSON → set Ollama node Base URL and model → Google Sheets header configuration (columns like leadSummary) → Telegram node permission settings.

Typical failures and fixes:

1. "Inquiry: undefined": Update the Parse LLM Response node code to ensure the inquiryMessage key is returned.
2. Chat interface displays JSON: Check that the Chat Response node returns the text key instead of response.
3. JSON parsing error: Disable "Format: JSON" in the Ollama node and lower the temperature to 0.1.

The system provides a zero-subscription-cost solution for small and medium-sized enterprises, replacing expensive commercial CRMs. Application scenarios: startups validating sales processes, freelancers managing client inquiries, small agencies unifying multi-channel leads, and tech teams integrating AI into existing workflows. It supports flexible expansion (e.g., email notifications, sentiment analysis, enterprise CRM integration).

The project combines modern AI and automation tools to build a fully functional and cost-controllable lead pipeline. Suggestions: Expand nodes according to business needs (e.g., email notifications, multi-model integration) or connect to enterprise-level CRM systems to enhance scalability.