AI Lead Qualifier — Score Any Domain List Against Your ICP

Stop manually eyeballing lead lists. Paste company domains, describe your ideal customer in plain language, and get back a 0-100 fit score for every domain — with concrete reasons and disqualifiers, plus a full company profile. No API keys, no scraping setup.

How it works

A frontier LLM reads each company's website and answers two questions:

1. What is this company? — name, description, industry, business model, products, contacts, socials
2. Should you contact them? — fit score against your ICP, with reasons grounded in what's actually on their site

Input

{
  "domains": ["stripe.com", "notion.so", "some-small-agency.com"],
  "icp_description": "B2B SaaS companies, 10-200 employees, selling to sales or marketing teams, based in North America or Europe."
}

Protocol prefixes and paths are stripped automatically; duplicates removed. Leave icp_description empty for enrichment only (no scoring).

Output

One dataset item per domain, sortable by fit_score:

{
  "domain": "notion.so",
  "ok": true,
  "fit_score": 78,
  "fit_reasons": [
    "B2B SaaS product with team collaboration focus",
    "Sells to knowledge-work teams including marketing"
  ],
  "disqualifiers": [
    "Larger than 200 employees"
  ],
  "company_name": "Notion",
  "description": "Notion provides a connected workspace for docs, wikis and projects...",
  "industry": "Software / productivity",
  "business_model": "b2b",
  "products_services": ["Docs", "Wikis", "Projects", "Notion AI"],
  "emails": [],
  "social_links": ["https://twitter.com/notionhq"],
  "tech_hints": []
}

Unreachable or broken sites produce { "domain": ..., "ok": false, "error": ... } and are not charged.

Pricing

Pay per event: one domain-qualified charge per successfully processed domain. Failed domains are free.

Use cases

• Score inbound signups before routing to sales
• Prioritize cold outreach lists — call the 80s first, skip the 20s
• Clean bought lead lists before importing to CRM
• Territory / market research with built-in qualification

Why scores you can trust

Scoring is grounded in site text only — the model must cite concrete fit_reasons and disqualifiers from the website, not guess. Companies whose sites say nothing relevant score low with an explanation, not a hallucinated high score.