<div align="center">

🤖 Multi-Agent System for Customer Outreach Campaigns with Tools

An autonomous AI-powered sales pipeline built with CrewAI, OpenAI & Live Web Search

---

---

💡 "What if your entire sales outreach process — research, profiling, and email writing — could run on autopilot, powered by AI agents?"

This project makes that a reality.

---

</div>

---

📌 Table of Contents

• 🌟 Project Overview
• ✨ Key Features
• 🏗️ System Architecture
• 🤖 The AI Agents
• 🛠️ Tools Used
• 📋 Tasks Breakdown
• ⚙️ Installation & Setup
• 🚀 How to Run
• 📊 Example Output
• 🧠 How It Works
• 📁 Project Structure
• 🔮 Future Improvements
• 👤 Author

---

🌟 Project Overview

This project implements a fully autonomous, multi-agent AI system designed to automate the entire customer outreach workflow — from deep company research to crafting hyper-personalized sales emails.

Built from scratch using CrewAI, it showcases how multiple AI agents can collaborate, delegate, and use real-world tools to complete complex business tasks without human intervention.

The system targets a specific lead (e.g. Google Cloud / Sundar Pichai), researches them using live web search, analyzes sentiment, and produces professional, personalized outreach emails — all autonomously.

---

✨ Key Features

| Feature | Description | |--------|-------------| | 🤖 Multi-Agent Architecture | Two specialized AI agents with distinct roles collaborating in a crew | | 🔍 Live Web Search | Real-time Google search via SerperDev API for up-to-date lead research | | 🧠 Persistent Memory | Agents share context and memory across tasks for coherent output | | 💬 Custom NLP Tool | Hand-built SentimentAnalysisTool to ensure positive communication tone | | 📂 File & Directory Tools | Reads local instruction files to guide agent behavior | | ✉️ Auto Email Generation | Produces professional, tailored outreach emails targeting specific decision-makers | | 🔧 Fully Extensible | Swap in any LLM, add new agents, or connect real sentiment models |

---

🏗️ System Architecture

┌─────────────────────────────────────────────────────────┐
│                        CREW AI                          │
│                                                         │
│   ┌─────────────────┐       ┌──────────────────────┐   │
│   │  Sales Rep      │──────▶│  Lead Sales Rep      │   │
│   │  Agent 🔍       │       │  Agent ✉️             │   │
│   └────────┬────────┘       └──────────┬───────────┘   │
│            │                           │               │
│   ┌────────▼────────┐       ┌──────────▼───────────┐   │
│   │ Lead Profiling  │       │ Personalized         │   │
│   │ Task            │       │ Outreach Task        │   │
│   └────────┬────────┘       └──────────┬───────────┘   │
│            │                           │               │
│   ┌────────▼──────────────────────────▼───────────┐   │
│   │     TOOLS: WebSearch | FileReader | Sentiment  │   │
│   └────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────┘
                          │
                          ▼
            📧 Personalized Outreach Campaign

---

🤖 The AI Agents

🔍 Agent 1 — Sales Representative

"The Researcher"

• Role: Scours the internet and internal files to build a detailed lead profile
• Goal: Identify high-value leads that match the ideal customer profile
• Skills: Data analysis, trend identification, strategic profiling
• Tools: DirectoryReadTool, FileReadTool, SerperDevTool

---

✉️ Agent 2 — Lead Sales Representative

"The Communicator"

• Role: Crafts personalized, compelling outreach based on research findings
• Goal: Nurture leads with tailored communications that convert interest into action
• Skills: Persuasive writing, sentiment analysis, brand tone matching
• Tools: SentimentAnalysisTool, SerperDevTool

---

🛠️ Tools Used

🔧 Built-in CrewAI Tools

| Tool | Purpose | |------|---------| | 📁 DirectoryReadTool | Reads local ./instructions folder for campaign guidelines | | 📄 FileReadTool | Opens and reads individual instruction files | | 🌐 SerperDevTool | Performs live Google searches for real-time company research |

🧪 Custom Tool — Built from Scratch

class SentimentAnalysisTool(BaseTool):
    name: str = "Sentiment Analysis Tool"
    description: str = (
        "Analyzes the sentiment of text to ensure 
         positive and engaging communication."
    )

    def _run(self, text: str) -> str:
        # Extensible: plug in HuggingFace, VADER, or any NLP model
        return "positive"

🔬 This custom tool was built by subclassing CrewAI's BaseTool — it can be extended with any real NLP model for production use.

---

📋 Tasks Breakdown

📊 Task 1 — Lead Profiling

Conducts deep research on the target company using all available tools. Produces a comprehensive report covering:

• Company background & industry position
• Key decision-makers & personnel
• Recent milestones & business developments
• Identified pain points & opportunities

📧 Task 2 — Personalized Outreach

Using the profiling report, crafts a series of tailored outreach emails that:

• Address the lead's specific recent milestones
• Match the company's culture and tone
• Connect solutions to the lead's goals
• Are validated for positive, professional sentiment

---

⚙️ Installation & Setup

Prerequisites

• Python 3.10+
• OpenAI API Key → Get one here
• Serper API Key → Get one here

Install Dependencies

pip install crewai==0.28.8 crewai_tools==0.1.6 langchain_community==0.0.29

Configure API Keys

import os
os.environ["OPENAI_API_KEY"] = "your-openai-key"
os.environ["OPENAI_MODEL_NAME"] = "gpt-3.5-turbo"   # or gpt-4-turbo
os.environ["SERPER_API_KEY"] = "your-serper-key"

⚠️ Never hardcode API keys. Use environment variables or a .env file.

---

🚀 How to Run

# 1. Clone the repository
git clone https://github.com/Pradeep-Kumar25th/multi-agent-customer-outreach.git
cd multi-agent-customer-outreach

# 2. Install dependencies
pip install -r requirements.txt

# 3. Open the notebook
jupyter notebook "Multi-Agent System for Customer Outreach Campaigns with Tools.ipynb"

# 4. Run all cells!

🎯 Customize Your Target Lead

inputs = {
    "lead_name": "Google Cloud",          # 🏢 Target company
    "industry": "Cloud Platform",          # 🏭 Industry
    "key_decision_maker": "Sundar Pichai", # 👤 Target person
    "position": "CEO",                     # 💼 Their role
    "milestone": "product launch"          # 🎯 Recent milestone
}

---

📊 Example Output

The system produces a full outreach campaign like:

Subject: Empowering Google Cloud's Next Chapter After Your Latest Product Launch

Dear Sundar,

Having closely followed Google Cloud's remarkable trajectory, particularly 
your recent product launch that has set new benchmarks in cloud innovation...

[Full personalized email crafted by AI agents]

---

🧠 How It Works

1. 📥  Input lead details (company, decision-maker, milestone)
         │
2. 🔍  Sales Rep Agent researches the company using web search + files
         │
3. 📝  Lead Profiling Task generates a comprehensive report
         │
4. ✉️  Lead Sales Rep Agent reads the report + writes outreach emails
         │
5. 😊  SentimentAnalysisTool validates tone is positive & professional
         │
6. 📤  Final personalized email campaign is output as Markdown

---

📁 Project Structure

📦 multi-agent-customer-outreach/
 ┣ 📓 Multi-Agent System for Customer Outreach Campaigns with Tools.ipynb
 ┣ 📁 instructions/
 ┃ ┗ 📄 campaign_guidelines.txt
 ┣ 📄 utils.py
 ┣ 📄 requirements.txt
 ┗ 📄 README.md

---

🔮 Future Improvements

• [ ] 🔁 Add a feedback loop agent to refine emails based on response rates
• [ ] 🧠 Integrate real sentiment model (HuggingFace / VADER)
• [ ] 📊 Add a reporting agent to track campaign performance
• [ ] 🗃️ Connect to CRM systems (HubSpot, Salesforce)
• [ ] 🌍 Multi-language outreach support
• [ ] 🖥️ Build a Streamlit UI for non-technical users

---

👤 Author

<div align="center">

Built from scratch with ❤️

If you found this project useful, please consider giving it a ⭐ — it helps others discover it!

---

Made with 🤖 CrewAI • 🧠 OpenAI • 🔍 SerperDev • 🐍 Python

</div>