Data Suitability Recipe

The Data Suitability recipe answers one question with evidence: did the data feeding this AI model or agent meet the documented standard? It does so by building an AI Use Case, the governance record that binds the model's input datastores and containers, the quality checks that define its standard, and its certification state. Once the record is certified, the recipe captures an attestation, an immutable, point-in-time evidence record you can hand to an auditor.

Unlike the other recipes, this one does not scan data or write tables. It reads the platform's existing metadata about the bound assets: quality scores across the eight dimensions, profiles and row counts, open anomalies, the check set and its outcomes, lineage edges, and a masking summary. Nothing in the evidence contains rows or values from your data. There is also no AI involved: every part of the record is assembled deterministically.

The recipe is resumable by design. Its progress is saved on the AI Use Case as a recipe run, so you can close it and continue later from the record list or from the AI Use Case page, on any device.

This Deep Dive covers:

• The two phases, Define the Use Case and Capture the Evidence, and the steps inside each
• The AI Use Case picker the recipe opens on, and how a run is saved and resumed
• What an AI Use Case is: certification lifecycle, who can do what, attestations, and audit obligations
• The entry points from the record list and the AI Use Case page
• Examples, best practices, and the permissions behind each step

What You Will Find Here