Breaking Down My AI Lead Scoring Architecture with n8n

Hello everyone,

I've been working on a project to solve a classic problem: lead response time. The goal was to move from manual sorting to instant, AI-powered qualification and routing.

I wanted to share the architecture I built and see what this community thinks.

The Problem: Hot leads get cold when they sit in an inbox for hours. Manual lead scoring is slow and inconsistent.

The Solution: A centralized system that automatically captures, enriches, scores, and routes every single lead instantly.

How It Works:

Capture & Enrich: A web form trigger captures the lead. The workflow then pulls the company URL from the form to enrich the profile with additional data.
AI Scoring (The Brain): The enriched data is sent to Google Gemini. It scores the lead from 1-100 based on criteria stored in a Pinecone vector database (our "rulebook").
Smart Routing: Hot leads (≥75) trigger an immediate CRM create, personalized email, and sales alert. Warm leads (25-74) are sent for sales review with a softer follow-up. Cold leads (<25) are logged for future nurturing, ensuring the sales team only focuses on the most promising opportunities.
Self-Learning: The system stays updated. New documents added to a Google Drive folder are automatically processed and added to the Pinecone knowledge base.

Key Tools: n8n (orchestration), Google Gemini (AI analysis), Pinecone (vector database for rules).

This has been an incredible learning experience in connecting AI models to real-world business automation. I'm really interested in this space.

I'd love to get your thoughts: