Curo - The Catalog Curator

Critical tables often sit without clear business context, which slows governance and weakens AI grounding. Curo identifies the tables that matter most, checks whether they already inherit a description, and routes targeted, non repetitive questions to the right steward, without overloading anyone who already has a full queue.

What you receive

How the recipe works (guided flow)

Step 1 — Baseline Importance Selection
Pulls core table metadata, row counts, and usage scores from the catalog, filtering out technical noise.

Step 2 — Recursive Downstream Impact Amplification
Traces lineage up to 5 levels deep to find dependent executive reports and boosts the importance score accordingly.

Step 3 — Multi-Factor Structural Classification
Classifies each table as Transactional, Master, Reference, or Configuration based on row volume, relationship density, and naming patterns.

Step 4 — Comprehensive Sensitive Data (PII) Discovery
Scans column names against a library of PII keywords to flag tables containing identity, contact, financial, or demographic data.

Step 5 — Final Criticality Scoring & Governance Prioritization
Combines value, risk, and classification into a single 0 to 100 criticality score and buckets tables into governance priority tiers.

Step 6 — Terms which Copy Business Description to Associated Data Object
Identifies business glossary terms configured to auto copy their description to linked tables.

Step 7 — Exclude Tables which Inherit Business Description through Associated Term
Removes tables that already receive a description automatically through one of those terms.

Step 8 — Critical Tables with Questions Already Generated through AI
Excludes tables that already have AI generated questions assigned, to avoid duplicate work.

Step 9 — Obtaining Business Descriptions
Joins in the existing business description for each remaining critical table.

Step 10 — Obtaining Stewards
Enriches each table with its assigned steward.

Step 11 — Steward Allocation Capping (Max 10 per Steward)
Caps each steward at 10 entries in this run to keep the workload balanced from the start.

Step 12 — Assignee-wise Unanswered AI Questions
Counts each steward's existing open AI generated questions and calculates how many more they can take on.

Step 13 — Capacity-Based Question Allocation for Stewards
Filters the table list against each steward's remaining capacity, preserving priority order throughout.

Step 14 — Generating Relevant Questions for Business Description Curation
Uses AI to generate up to 3 distinct, non repetitive questions per table based on business purpose, stakeholders, usage, business rules, KPIs, and regulatory implications.

Step 15 — Question Set for Business Context Enrichment
Reorders and renames the final columns into a clean, steward-ready dataset.

Step 16 — Configuring Action to Assign Questions
Posts each question to the relevant table's question wall and assigns it to the correct steward.

Sample insights (illustrative)

Make sure the following ingredients are available in your workspace:

• Connected datasets are crawled and profiled