WaterfallReach
An AI agent that descends a five-tier waterfall — so you only pay for as much waterfall as the row needs.
The fastest path to a verified work email is not always the most expensive one.
Single-vendor contact enrichment is a black box: one API call, one charge, one verdict. But most contacts can be found for free if you know where to look. We built an AI agent that knows where to look — and only pays when it has to.
┌───────────────────────────────────┐
│ INPUT: linkedin · email · name │
└────────────────┬──────────────────┘
│
▼ ─────── TIER 0 — free signals (~20+)
│
▼ ─────── TIER 1 — self-hosted SMTP
│
▼ ─────── TIER 2 — LinkedIn router + AI
│
▼ ─────── TIER 3 — paid finders
│
▼ ─────── TIER 4 — verification
│
┌──────┴───────────────────────────────┐
│ OUTPUT: verified email · T-class │
└──────────────────────────────────────┘
▲
(exit on first verified match)
The way most teams enrich contact data today is to wire a single vendor — Apollo, ZoomInfo, Cognism, Lusha — into their pipeline and charge per row. The vendor does its best work and bills you whether the answer was already free, already verifiable from a public record, or genuinely required a paid lookup. You pay the same regardless.
We thought this was strange. Permutations of common email patterns work on most domains. MX fingerprints reveal which mail systems will accept a guess. IRS Form 990 lists every officer of every non-profit hospital in the country. The U.S. NPPES registry has every licensed clinician. SEC EDGAR has every public-company executive. None of this requires a paid API. It just requires patience.
"You pay for as much waterfall as the row needs — not for the whole vendor."
An AI agent that knows where to look
The piece that makes a free-first approach work in production is judgment. Twenty-plus weak signals are not a verified email; they're a probability distribution. So our AI agent does two things conventional enrichment cannot:
- Fuses signals into a ranked candidate. The AI judges every signal it gathers — permutations, public records, LinkedIn cache, MX fingerprint — and produces one ordered list. The top candidate gets sent to SMTP verification. If it lands, the cascade exits at Tier 0.
- Learns each domain. Once the AI verifies a pattern for acme.com, every future request on that domain skips the expensive tiers automatically. The system gets cheaper as it gets used.
The waterfall has five tiers
Tier 0 is free signals. Tier 1 is self-hosted SMTP validation that detects security gateways (Proofpoint, Mimecast, Barracuda, Cisco) which would otherwise produce false-positive verifications. Tier 2 is a LinkedIn router across six providers, with an AI ranker that picks the best result. Tier 3 is paid finders — Prospeo, Findymail, Dropcontact — but only fired when nothing earlier resolves. Tier 4 is verification with an AI trust-score gate that cuts paid-verifier spend before any paid verifier is called.
Every tier carries a T0–T3 legal-risk class. Customers cap the ceiling. GDPR posture happens at configuration, not in code.
Bring a list. See where it exits.
Pilot access is open. Drop a CSV of LinkedIn URLs, domains, or names — we'll run the cascade and return a report showing each row's exit tier, source, and cost.
WaterfallReach, est. MMXXVI