Data Dictionary Maintainer

The Challenge

Data dictionaries become stale quickly. Schema changes happen faster than documentation updates, leaving analysts guessing about field meanings and data lineage.

The AI Desk Solution

AI Desk monitors schema changes and updates data dictionaries with inferred descriptions and lineage.

The Workflow

Step 1: Schema Monitoring

Trigger: Schema change detected
Sources: Database metadata, dbt, existing docs

Step 2: Documentation Update

• Field description inference
• Lineage tracking
• Usage pattern analysis

Step 3: Updated Dictionary

📖 Data Dictionary Update

SCHEMA: analytics.customers

NEW FIELDS DETECTED (3)

customer_health_score
├── Type: INTEGER
├── Added: Apr 24, 2026
├── Source: Derived (health_model.sql)
├── Description: Composite score (0-100)
│   based on usage, support, and
│   engagement signals
├── Updated by: Jamie Chen
└── Related: customer_risk_level

last_nps_response
├── Type: TIMESTAMP
├── Added: Apr 24, 2026
├── Source: nps_surveys table
├── Description: Most recent NPS survey
│   completion timestamp
└── Related: nps_score, nps_feedback

expansion_potential
├── Type: VARCHAR(20)
├── Added: Apr 24, 2026
├── Values: low, medium, high
├── Description: ML-predicted likelihood
│   of account expansion in next 90 days
└── Related: expansion_revenue

MODIFIED FIELDS (1)

customer_segment
├── Change: Added new value 'strategic'
├── Previous values: smb, mid_market, enterprise
├── New values: + strategic
└── Reason: New tier for $500K+ accounts

DEPRECATED FIELDS (1)

legacy_account_id
├── Status: Deprecated
├── Replacement: account_uuid
├── Removal date: Jul 1, 2026
└── Migration: Complete

LINEAGE UPDATED
├── customer_health_score ← health_model
├── health_model ← [usage, support, engagement]
└── Diagram: [View in catalog]

Value Proposition

• Time Saved: 4 hours per update cycle
• Always Current: Synced with schema
• Self-Documenting: Descriptions inferred