CrewAI Sales Leads Agent

What is this project?

An AI-powered sales prospecting assistant that discovers and enriches potential leads based on your ideal customer profile. It automates web research, enrichment, and basic verification to produce a structured, deduplicated list of prospects you can export.

Why did I build it?

To experiment with multi-agent collaboration using CrewAI and see how far agents can go in automating real, end-to-end workflows. Lead generation is a practical domain where research, enrichment, and filtering come together—and it’s ideal for testing agent handoffs.

How does it work?

1. You define a niche (industry, role, geography, keywords)
2. A research agent uses web search to find candidate companies and contacts
3. An enrichment agent visits sources (sites/profiles) to extract structured fields
4. A verifier agent dedupes and prunes low-signal entries
5. Results are presented in a Streamlit UI with CSV export

Tech Stack

• Python
• CrewAI
• Serper API (web search)
• Streamlit

What were the biggest challenges?

• Coordinating agent boundaries so they collaborate instead of looping
• Normalizing messy web data into consistent, comparable fields
• Handling rate limits and partial failures gracefully without losing progress
• Deduplication heuristics (domain + name + role) to reduce repeated entries

How can you use it or build on this?

• Try a niche (e.g., “AI startups in healthcare, CTO/Head of ML, US”)
• Export the CSV and feed into your CRM or outreach tool
• Extend with email finding/verifying, or plug into a CRM API
• Code: https://github.com/Folken2/lead-generator-ai

What’s next?

• Add more sources and structured enrichment steps
• Integrate basic contact discovery with verification
• One-click export to HubSpot or Salesforce

In the age of AI, stay curious!