Pruvz builds a business evidence layer for AI agents: it captures the decision-time context and policy behind high-impact agent actions such as refunds, claims and account changes, independently checks the outcome against systems of record like billing and CRM, and routes mismatches to human review. The product runs end to end against Stripe and HubSpot test environments, with customer pilots and a founding design-partner program open.