Lead Sourcing Pipeline
A standalone service that sources cold leads for the freight operator's CRM entirely from public data. It bulk-imports the tax authority's monthly release of the national company registry, profiles which industries and regions already produce profitable clients, and queries lookalikes against its own database — so finding a candidate costs nothing.
Active Development
Candidate discovery costs nothing per lead. The full national establishment registry is imported and held locally, so matching a lookalike profile of industry codes and regions is a SQL query against your own database rather than a metered API call for every company considered.
The registry publishes no website, and almost everything downstream needs one. The resolver derives it in four tiers ordered by cost: the domain of the e-mail the company declared to the tax authority, a site already returned by a call that was paid for anyway, guessed spellings resolved in parallel across nine TLDs, and only then a single grounded AI search.
Every tier's answer passes the same corroboration gate. A page loading proves the domain exists, not that it belongs to this company — it must also mention the registration number, two distinctive words of the name, or the municipality. Words too common across the industry are excluded from that test, since matching on them would corroborate any competitor equally well.
Cost is an architectural invariant rather than a configuration value. A company reaches the single paid contact lookup only after a human approved it and its domain was independently corroborated; a domain that fails only corroboration is still stored and still shown to a rep to judge by hand.
The entitlement check fails closed. Before anything runs, the pipeline asks the platform whether the module is enabled for that tenant, and treats a timeout, a repeated server error, or a misconfigured secret as disabled rather than proceeding on an unconfirmed permission.
The service is deliberately standalone — its own database, deploy, and license — integrating with the CRM only through its public REST API. That keeps the upstream CRM fork cleanly mergeable and keeps a bulk-ETL workload out of a monorepo that should not be carrying one.