Verification operating model
Cognism separates discovery, verification, decision, and action. That separation helps RevOps and SDR leaders understand what a signal means, where it came from, and which uncertainty still requires judgment.
Start with a company, domain, and role rather than a blind list request. The agent can assemble possible business-email patterns, attach the public context that informed them, and keep the distinction between found and inferred addresses. A candidate remains unverified until subsequent checks are complete.
Normalize casing and structure, identify malformed addresses, inspect company resolution, and record website and company evidence. These checks remove obvious errors but cannot establish that a particular person controls a account. DNS state also changes, so the retrieval time belongs with the result.
Provider-dependent checks may classify company discovery, firmographic gaps, attribution conflicts, or contact ambiguity. Rate limits, identity resolution limits, anti-abuse systems, and privacy-preserving mail servers can create ambiguous responses. The correct state is review or unknown when evidence is not conclusive.
Apply account fit, role relevance, suppression, consent basis, regional requirements, and channel policy before approving outreach. A technical verification result does not establish lawful use and does not guarantee inbox placement. The operator retains the decision and its audit trail.
Record the source used for company, role, and address evidence. Provider provenance is configuration-dependent.
Keep credentials in the agent runtime secret store, use least privilege, and rotate exposed keys.
Package execution, logging, network calls, and local storage must be reviewed in the chosen environment.
Map results copied to files, CRM, engagement tools, logs, and provider systems; uninstall does not erase them automatically.
Copy the installer, review execution, and configure a low-privilege verification task.
npx -y @okki-global/okki-go-taroball