The Problem with Traditional Lead Gen
Most lead generation tools work the same way: you get a database dump, filter by industry and company size, and hope the data isn't stale. Spoiler — it usually is.
We wanted something different. Instead of selling static databases, we wanted to search the web in real time and let AI decide which companies match your criteria.
How LIZZY Works
LIZZY is LeadScoutr's conversational AI. When you describe your ideal customer — "B2B SaaS companies in the Netherlands with 20-100 employees that use HubSpot" — here's what happens under the hood:
1. Intent Parsing
LIZZY uses a language model to parse your natural language request into structured search parameters. Industry, location, company size, tech stack, funding stage — she extracts all of it from a single sentence.
2. Web Search
Using SERP APIs, LIZZY generates search queries and retrieves relevant company listings. She doesn't just search once — she generates multiple query variations to maximize coverage.
3. Company Scraping
For each search result, we use Firecrawl to visit the company website and extract structured data. Company description, employee count, tech stack indicators, contact information, and recent news.
4. Qualification Scoring
Every company gets scored against your ideal customer profile. The scoring model considers:
- Firmographic fit — size, industry, location
- Tech stack alignment — do they use the tools you integrate with?
- Growth signals — hiring, funding, product launches
- Engagement potential — how likely are they to need your product?
5. People Enrichment
Once a company is qualified, LeadScoutr pulls decision-maker profiles from People Data Labs. Names, job titles, verified email addresses, and LinkedIn URLs — everything you need to start outreach.
The Stack
LeadScoutr is built with:
- Next.js 15 for the application layer
- Prisma for database management
- Anthropic Claude for natural language understanding
- SERP APIs for real-time web search
- Firecrawl for web scraping and data extraction
- People Data Labs for contact enrichment
What We Learned
Building an AI-powered search engine taught us a few things:
-
Real-time data beats databases. A company that raised funding last week won't show up in a monthly database refresh. Real-time search catches it immediately.
-
Natural language input reduces friction. Users don't want to learn Boolean search syntax. They want to describe what they're looking for and get results.
-
Scoring needs transparency. Users trust AI recommendations more when they can see why a company scored high or low. We show the breakdown.
What's Next for LIZZY
We're working on multi-turn conversation support, saved search templates, and automated pipeline updates. The goal is to make LIZZY your always-on sales research assistant.