A score alone is not enough. Understand where information came from, when it was collected, what was inferred, and who made the decision.
Source statements describe what public material says. AI inferences are possible interpretations, not confirmed facts. User-provided information comes from your team, without invented citations. Unknown or unverified means there is not enough evidence. Stale information needs checking again. Samples are fictional, not real research.
Check source, collection, and assessment dates separately, with the relevant versions. A missing source date is unknown; collection time must not stand in for it.
AI structures source evidence; ICP rules calculate fit. Human review should examine evidence, unknowns, and conflicting information—not just accept a score.
Product PDFs are stored privately for the workspace and shared with the configured AI provider for extraction. Do not upload passwords, internal costs, or sensitive material. A formal data notice still needs to confirm processing, retention, and deletion details.
Real workspaces use membership permissions. Research has a separate estimate and approval; recording a review does not authorize sending. When task or billing status is uncertain, check the status before retrying.
Live contact discovery, sending, automatic booking, and CRM writes are not available. Simulations do not represent customer acceptance, opportunities, or revenue.