ProvAI
Built for clinical AI governance from day one.
ProvAI is a platform designed to support AI Governance for the clinical development lifecycle. It brings together zLab for building and validating AI Assets and SherpAI for controlled AI execution within one secure, audit-ready environment. Built around Programmer and Gatekeeper governance, ProvAI supports controlled development, approval, execution, and traceability, with 21 CFR Part 11 and GxP-aligned controls. With BYOL support and integration with existing systems and workflows, ProvAI enables organizations to adopt AI at scale without disrupting established processes.
zLab
Controlled Build Environment: Where AI Assets are built, validated, and approved
Governed Authoring
Build, test, validate, and document AI Assets in RStudio or VS Code while applying your organization’s own verification and approval processes within a controlled, governed development environment.
Data Policy Enforcement
Enforce organization-defined data policies before files enter development. Automated scanning, quarantine, and approval workflows help protect sensitive data, with all activity captured in a traceable audit trail designed to support 21 CFR Part 11 requirements.
AI Asset Library
Brings approved AI Assets together in a centralized library with version history, validation results, and approval records. Published Assets remain controlled and traceable, with separate CIMS-managed and organization-specific libraries supporting secure, reusable AI across the suite.
SherpAI
Production Execution + Audit Governance: Where approved AI Assets run in governed production
GxP Compliant
Runs approved AI Assets in a controlled, auditable environment designed to support GxP requirements. Role-based access and comprehensive audit trails help maintain control, accountability, and traceability throughout AI execution.
Approval Gates
Ensures AI-generated outputs pass defined approval gates before moving forward. Configurable approval policies give organizations control over how AI outputs are reviewed and approved.
Deidentified Source Data
Protect sensitive data before AI execution. DeID secures sponsor-selected information during processing, while ReID restores authorized outputs when needed.
