01
Interpretable
Can the data be correctly understood by AI consumers?
The Shift
Humans resolve ambiguity. AI encodes it. If meaning isn't explicit in the data, model performance suffers.
Requirements
What must be true about the data itself.
- Data meaning is explicit and machine-readable
- Field semantics are consistent across contexts
- Business logic is documented and accessible
- Data origin and transformations are traceable
- Schema contracts are enforced (types, nullability, valid values)
Capabilities
What your infrastructure must support.
Semantic layer / metrics layer
Business definitions that disambiguate terms across contexts
Data catalog / business glossary
Searchable, queryable field definitions and documentation
Column-level lineage
Trace any field back to its source and through transformations
Metadata APIs
Programmatic access to definitions, lineage, and context