An organization is scaling multiple AI initiatives across various departments.

The scenario highlights a breakdown in data lineage tracking across multiple transformations , which impacts auditability and transparency. The key issue is not data quality but the inability to trace how data evolves from its original source through the pipeline.

In CAIPM-aligned data architecture, lineage tracking must begin at the earliest point where data enters the AI pipeline , specifically during the stage where data is ingested and validated. This is where:

Data is first standardized and checked for quality

Metadata and lineage tracking mechanisms are initialized

Each transformation step can be recorded and linked back to the source

If lineage tracking is not established at this early stage, it becomes difficult or impossible to reconstruct data flows later, especially after multiple transformations and feature engineering steps.

Other options are less appropriate:

Model consumption stage occurs too late; lineage should already be established

Curated datasets stage organizes data but relies on prior lineage tracking

Data origin stage identifies the source but does not ensure tracking across transformations

CAIPM emphasizes that traceability must be built into the data pipeline from ingestion onward , ensuring that every transformation is auditable and linked to its origin.

Therefore, the correct answer is Where data is first validated and lineage tracking begins , as this is the critical point to establish transparency and auditability controls.

=========