AI Lead Qualification System

An end-to-end automated pipeline that captures leads, scores them with an LLM, routes them by priority, and notifies your sales team — no manual triage required.

---

Table of Contents

• Overview
• Demo
• Problem
• Solution
• Architecture
• How It Works
• Key Features
• Output Previews
• Tech Stack
• Project Structure
• Getting Started
• Warm Lead Monitoring
• Data Storage Strategy
• Future Improvements
• License

---

Overview

The AI Lead Qualification System is a fully automated pipeline designed to eliminate manual lead triage. Leads submitted through a web form are scored by an LLM, classified into priority tiers, stored in a CRM, and routed to the sales team via Slack — all in real time.

Supports both cloud-based (Groq) and local LLM deployments (Ollama), making it adaptable for cost-sensitive or air-gapped environments.

---

Demo

Click to watch full system demo

End-to-end walkthrough: form submission → LLM scoring → CRM update → Slack alerts

---

Problem

Manual lead qualification is slow and inconsistent. Sales teams waste time triaging low-intent submissions while high-priority leads sit unattended. Without a structured scoring layer, follow-up is delayed and revenue opportunities are lost.

---

Solution

An AI-driven pipeline that evaluates lead intent using natural language understanding, assigns a score from 1–100, and routes leads automatically based on priority — with zero manual intervention.

---

Architecture

• Upper flow — Real-time lead ingestion, LLM scoring, and routing
• Lower flow — Scheduled warm lead monitoring and follow-up
• A Notified flag ensures idempotent notifications and prevents duplicate alerts

---

How It Works

1. Lead submits the intake form → workflow triggers immediately
2. Submitted data is normalized and validated
3. A confirmation email is sent to the lead
4. The LLM scores the lead (1–100) and returns:
   • score — numerical priority rating
   • reasoning — natural language explanation
   • recommended_action — suggested next step
5. Duplicate check is performed via email address
6. Lead is classified and routed:

Score	Tier	Action
≥ 80	🔴 Hot	Instant Slack alert + CRM entry
50–79	🟡 Warm	CRM entry + scheduled follow-up
< 50	🔵 Cold	Logged to Google Sheets

7. All events are appended to the activity log
8. Notified flag is updated to prevent re-alerting

---

Key Features

• LLM-based scoring — Evaluates purchase intent, urgency, and service relevance
• Real-time + scheduled workflows — Instant processing with periodic warm lead monitoring
• Idempotent notifications — Notified flag prevents duplicate Slack alerts
• Structured Slack alerts — Actionable messages with score, reasoning, and next steps
• Email deduplication — Detects and updates existing records on resubmission
• Activity logging — Append-only Google Sheets log for auditing and analytics
• Robust LLM parsing — Handles malformed model outputs with fallback logic
• Local deployment support — Compatible with Ollama for zero-API-cost operation

---

Output Previews

Lead Intake Form

User-facing form for capturing lead information.

---

Confirmation Email

Automated email dispatched immediately after form submission.

---

Hot Lead — Slack Notification

Real-time alert for high-intent leads with score, insights, and recommended action.

---

Warm Lead — Slack Summary

Periodic digest of unnotified warm leads, sorted by score.

---

CRM — Airtable

Centralized lead storage with stage and notification tracking.

---

Activity Log — Google Sheets

Append-only interaction log for audit trails and downstream analytics.

---

Tech Stack

Tool	Role
n8n	Workflow orchestration
LLaMA 3.1 8B (Groq)	Lead scoring via LLM
Airtable	CRM and lead storage
Google Sheets	Append-only activity log
Tally	Lead intake form
Slack	Sales team notifications
SMTP	Confirmation email delivery

---

Project Structure

ai-lead-qualification-system/
│
├── README.md
├── docs/
│   └── SYSTEM_DESIGN.md
│
├── assets/
│   ├── system-architecture.png
│   ├── slack-hot.png
│   ├── slack-warm.png
│   ├── airtable.png
│   ├── form.png
│   ├── sheets.png
│   └── confirmation_mail.png
│   └── demo-thumbnail.png
│
├── snippets/
│   ├── llm-prompt.txt
│   ├── parser.js
│   ├── slack-format-hot.js
│   └── slack-format-warm.js
│
├── demo/
│   └── demo.mp4
│
└── LICENSE

---

Getting Started

Prerequisites

• An active n8n instance (self-hosted or cloud)
• A Groq API key or a local Ollama instance
• Airtable, Google Sheets, Slack, and SMTP credentials

Environment Variables

Configure the following in your n8n environment or .env file:

SLACK_HOT_WEBHOOK=
SLACK_WARM_WEBHOOK=
GROQ_API_KEY=
AIRTABLE_TOKEN=
SMTP_HOST=
SMTP_USER=
SMTP_PASS=

For local LLM deployment, replace the Groq HTTP node with an Ollama request node pointing to your local endpoint.

---

Warm Lead Monitoring

A scheduled workflow runs periodically to catch warm leads that haven't been followed up:

1. Fetches records where Stage = Warm and Notified = false
2. Sorts leads by score (descending)
3. Sends a structured Slack summary
4. Updates Notified = true for all included leads

If no qualifying leads are found, a fallback message confirms the workflow ran successfully.

---

Data Storage Strategy

Store	Purpose
Airtable	Live CRM — current lead state and stage
Google Sheets	Append-only log — full interaction history

This dual-store approach enables full lifecycle tracking, non-destructive auditing, and analytics-ready exports without polluting the CRM.

---

Future Improvements

• Replace polling with event-driven triggers
• Add retry logic for API/LLM failures
• Implement LLM fallback model
• Lead enrichment via LinkedIn, Clearbit or online sources.
• Analytics dashboard (scores, conversion rates, tier distribution)
• Cold lead re-engagement workflow

---

License

This project is licensed under the MIT License.