AI Readiness Evaluation

Text-to-SQL · RAG readiness · Vectorization candidates · Label consistency · Join complexity

AI Confidence Score

/ 100

Fair Readiness

Significant blockers to resolve

Text-to-SQL Score

/ 100 · Moderate

Vectorization Score

18 candidates

Join Complexity

High

avg 4.2 joins

Data Completeness

84%

Non-null avg

Label Consistency

73%

Naming patterns

AI Blockers

Must resolve

AI Readiness Radar

Dimension Scores

Data Completeness84%

High NULL rates in 14 columns impacting model training.

Schema Clarity (Text-to-SQL)64%

Ambiguous column names (e.g., val, flag, type) confuse LLM query generation.

Vectorization Potential71%

18 text columns suitable for embedding; 6 need cleaning.

Join Integrity52%

Deep join chains (4+ hops) make context retrieval unreliable for RAG.

Label Consistency73%

Mixed snake_case/camelCase naming in analytics schema.

RAG Suitability48%

Schema normalization issues cause poor chunk retrieval quality.

AI Blockers — Must Resolve Before AI Integration

5 blockers

Ambiguous column names prevent reliable Text-to-SQL

38 columns with generic names (val, flag, type, data, info) across 12 tables. LLMs cannot reliably infer semantics → wrong SQL generation.

Fix

Missing FK constraints break RAG relationship traversal

4 implied relationships not declared as FKs. AI agents cannot reliably discover join paths for context retrieval chains.

Fix

High NULL rates in label columns

14 columns with >40% NULL values in tables used for classification training. This biases model outputs.

View

Mixed naming conventions degrade prompt injection accuracy

analytics schema uses camelCase; public schema uses snake_case. Cross-schema queries are error-prone for AI generation.

View

Text columns not normalized for embedding

6 text columns contain HTML tags, control characters, or inconsistent encoding that will degrade vector search quality.

View

Top Vectorization Candidates

Table	Column	Data Type	Use Case
products	description	TEXT	Semantic search
support_tickets	body	TEXT	RAG retrieval
articles	content	TEXT	Knowledge base
reviews	review_text	TEXT	Sentiment
employees	bio	TEXT	HR assistant

Suggested Restructuring for AI

Add embeddings companion table

Create product_embeddings(product_id, vector VECTOR(1536), model VARCHAR) for pgvector integration.

Normalize column naming to snake_case

Rename 22 camelCase columns in analytics schema to snake_case for consistent LLM context injection.

Add semantic column comments

COMMENT ON COLUMN for 38 ambiguous columns to provide schema context to LLMs and improve Text-to-SQL accuracy by ~40%.

Create schema_metadata view

Expose table descriptions, relationships, and data contracts as a queryable view for AI context injection.