Lead Discovery: How it Works

Lead Discovery: How it Works

wrrk Lead Discovery turns a single sentence into hundreds of qualified leads, anywhere in the world. Type "Dental clinics in Berlin" or "Indian SaaS founders raising Series A" and an agent swarm fans out across 8 sources, enriches each candidate with phone, email, website, and signals, and streams cards into your workspace as it works.

The three intelligence layers

1. ICP Router: Reads your prompt, classifies the ideal-customer-profile shape (tech-founder, smb-local, b2b-services, b2g, enterprise, tech-engineer, or mixed), and picks the right combination of sources for the geography. You don't pick sources manually, the system picks them based on what your prompt is actually asking for.

2. Source Discovery: For non-English geographies (Lithuania, Brazil, Japan, Nigeria, etc.) the system identifies the right local directories and news outlets at runtime. The first query for a new country takes ~10s longer; every subsequent query for that country is instant via cache.

3. Cascade Resolver: Every news article, post, or signal flows through a Bedrock-powered extractor that pulls company → website → leadership → emails into structured candidate rows. One funding announcement becomes 5-30 enriched leads.

The 8 sources

What's on a lead card?

Every candidate ships with: name, title, company, source badge, qualification score, decision-power score, why-match (one-line judge rationale), per-candidate signals (e.g. "recent funding announcement", "actively building", "phone listed"), and as much enrichment as we could find, email (with confidence %), phone (click-to-call), website, address, LinkedIn, Twitter, Instagram, Facebook, YouTube.

Geographic coverage

Any country, any city. OSM uses OpenStreetMap's Nominatim to geocode any place name on Earth. The Directory hunter discovers country-specific public directories at runtime. News-RSS pulls from 22 globally-distributed feeds covering India, US, EU, SE Asia, China, Africa, and LATAM. Tested geographies that consistently produce 50+ candidates per query: India, Germany, UK, France, Italy, Spain, Lithuania, Australia, Canada, Brazil, Mexico, Singapore, Japan, Nigeria.

What it isn't

This is not a static database lookup like Apollo or ZoomInfo. Each query runs a fresh agent swarm across live public sources at the moment you ask. The data is therefore fresh (signals from the last 7-30 days), but query latency is 60-180 seconds, not milliseconds. For repeat lookups, candidates are cached in your CRM after you accept them.