Learn / Glossary
AI Governance Glossary
Key terms and definitions for enterprise AI governance, work verification, and accountability infrastructure.
- Trusted Work Unit
- A tamper-evident, cryptographically sealed record of a completed task that captures who or what performed the work, the full evidence chain, and the verified outcome. Trusted Work Units are the foundational data structure of the Veratrace platform, enabling deterministic metering, attribution, and audit across human and AI actors.
- AI Work Verification
- The practice of confirming that tasks performed by AI systems actually occurred, produced the claimed outputs, and met the standards required by the organization. Unlike logging, verification produces tamper-evident records with cryptographic integrity that can withstand regulatory scrutiny.
- AI Agent Audit Trail
- A complete, immutable record of every meaningful action an AI agent takes during task execution — including inputs received, tools invoked, decisions made, and outcomes produced. An effective audit trail is sealed into verifiable records that auditors can independently validate without relying on the AI vendor's own logging.
- AI Work Attribution
- The process of determining and quantifying the contributions of human agents, AI systems, and automated processes within a single task. Attribution uses evidence-based analysis — including time weighting, edit significance, and rework detection — to calculate defensible contribution percentages for each actor.
- Automation Reconciliation
- The practice of comparing vendor-reported AI usage and billing against independently captured work records. Reconciliation identifies discrepancies between what vendors charge and what enterprises can verify, including retried calls billed as separate events and AI-generated work that required full human rework.
- AI Governance Infrastructure
- The operational systems and controls required to govern AI at enterprise scale — beyond policy documents and ethics committees. Governance infrastructure captures evidence, enforces controls, and produces audit-ready records as a byproduct of normal operations, bridging the gap between policy intent and operational reality.