How is the traceability of AI results ensured?
amaise ensures complete traceability of all AI-generated results through multiple mechanisms:
Audit trail: A dedicated audit service records all data operations with user, action, entity, IP address, timestamp, and access groups.
Pipeline tracking: Pipeline state transitions are logged and link processing stages to specific documents.
Token tracking: LLM token usage is recorded per pipeline stage, enabling tracking of AI usage.
Model pinning: Model versions are defined per environment in the infrastructure configuration and are traceable.
Log retention: All service logs are retained for 365 days.
User actions: All user actions (accepting, modifying, rejecting AI results) are recorded in the audit trail.
AI results are presented in the context of the source document, allowing users to review and verify the basis of each result.