B2B Lead Prospecting Pipeline

An AI-powered B2B lead prospecting pipeline built with LangGraph that automates company enrichment, contact discovery, and lead qualification against a configurable Ideal Customer Profile (ICP).

The system fetches leads from your CRM, enriches each company with firmographic data and recent business signals, discovers decision makers, scores the lead on a 0–100 scale with an A/B/C grade, and syncs the results back to your CRM — all automatically.

Features

Multi-CRM Integration

• Connect to HubSpot, Airtable, Google Sheets, or add your own CRM by extending the base class.

Automated Company Enrichment

• Website Scraping: Extract industry, employee count, tech stack, and location from the company website.
• Recent Signals: Fetch and analyze recent news for funding rounds, hiring surges, product launches, and partnerships.
• No Paid APIs: All enrichment is inferred from publicly available web data via search and scraping.

Contact Discovery

• Automatically find decision makers (C-suite, VPs, Directors) using web and LinkedIn search.
• Extract name, title, and email when available.

Lead Qualification & Scoring

• LLM-based scoring: Evaluates each lead against your ICP across 5 dimensions (industry fit, size fit, funding fit, signal strength, tech fit) on a 0–100 scale.
• Rule-based fallback: If the LLM call fails, a deterministic scorer produces the same output structure.
• Grade assignment: A (≥70), B (≥45), C (<45) with disqualification reasons and recommended outreach angles.

CRM Sync

• All scoring results, grades, qualification status, pain points, and outreach angles are written back to your CRM automatically.

System Workflow

get_new_leads → check_remaining_leads → enrich_company → discover_contacts → qualify_lead → sync_to_crm → (loop back)

1. Fetch Leads: Pull new leads from your CRM.
2. Enrich Company: Scrape the company website and recent news to populate firmographic data and business signals.
3. Discover Contacts: Search for decision makers at the company.
4. Qualify Lead: Score against your ICP using an LLM (with rule-based fallback). Produces score, grade, qualified status, pain points, and outreach angle.
5. Sync to CRM: Write all results back to the CRM record.
6. Loop: Process the next lead until none remain.

Configuration

ICP Config (config.yaml)

Define your Ideal Customer Profile at the project root:

target_industries: ["SaaS", "Data SaaS"]
min_employees: 50
max_employees: 600
funding_stages: ["Series B", "Series C"]
target_locations: ["United States"]
target_tech: ["Stripe", "AWS", "GCP"]

The qualification prompt and rule-based scorer both use these values.

Integration with APIs

• Airtable CRM: Sign up and create a contacts table with relevant fields.
• HubSpot CRM: Sign up, create a private app, and get your API key.
• Google Searches: Perform web searches using the Serper API. Get your API key here.
• LLM: Uses Google Gemini Flash via langchain_google_genai. Get your API key here.
• Google APIs (optional): Used for Google Sheets CRM and Gmail. Follow this guide.

Tech Stack

• LangGraph: State machine framework for the prospecting workflow.
• LangChain: LLM abstraction and tooling.
• Google Gemini Flash: Primary LLM for enrichment and qualification.
• PyYAML: ICP configuration loading.

How to Run

Prerequisites

• Python 3.9+
• Google Gemini API key
• Serper API key
• CRM credentials (check .env.example)

Setup

1. Clone the repository:

   git clone https://github.com/kaymen99/sales-outreach-automation-langgraph.git
   cd sales-outreach-automation-langgraph

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install dependencies:

   pip install -r requirements.txt

4. Set up environment variables:

   cp .env.example .env

   Open .env and add your API keys.

5. Configure your ICP:

   Edit config.yaml with your target industries, employee range, funding stages, locations, and tech stack.

---

Start the Pipeline

python main.py

The system connects to your CRM, fetches new leads, enriches each company, discovers contacts, qualifies leads, and syncs results back.

---

Customizing

• Change your ICP: Edit config.yaml — no code changes needed.
• Add a custom CRM: Extend LeadLoaderBase in src/tools/leads_loader/lead_loader_base.py.
• Adjust scoring: Modify LEAD_QUALIFICATION_PROMPT in src/prompts.py or the rule-based fallback in src/scorer.py.
• Add enrichment sources: Extend src/tools/enrichment.py.

Project Structure

├── config.yaml                 # ICP configuration
├── main.py                     # Entry point
├── requirements.txt
├── src/
│   ├── graph.py                # LangGraph workflow wiring
│   ├── nodes.py                # All pipeline node logic
│   ├── prompts.py              # LLM prompts
│   ├── scorer.py               # Rule-based scoring fallback
│   ├── state.py                # Graph state schema
│   ├── utils.py                # LLM invocation, Google auth
│   └── tools/
│       ├── enrichment.py       # Company data enrichment
│       ├── contact_discovery.py # Decision maker discovery
│       ├── base/
│       │   ├── search_tools.py       # Google search & news
│       │   ├── markdown_scraper_tool.py
│       │   ├── gmail_tools.py
│       │   └── linkedin_tools.py
│       └── leads_loader/
│           ├── lead_loader_base.py
│           ├── airtable.py
│           ├── google_sheets.py
│           └── hubspot.py

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

Contact

For questions or suggestions, contact aymenMir1001@gmail.com.