Automating data classification in Amazon SageMaker Catalog using an AI agent

This article explains how Amazon SageMaker Catalog's AI agent automates data classification using business glossaries and Amazon Bedrock language models.

• AI agent analyzes table metadata and schema to suggest relevant business glossary terms automatically
• Reduces manual tagging effort and improves metadata consistency across organizations
• Agent uses reasoning-driven approach: reviews context, searches catalog, evaluates results, refines strategy
• Suggestions include functional terms and sensitive data classifications like PII and PHI
• Data producers review and accept or modify suggestions before publishing assets
• Integrated directly into publish workflow with no separate ETL processes required
• Requires well-defined, consistent business glossaries to generate accurate recommendations
• Enables improved data discovery through standardized terminology and filtering by terms

SageMaker Catalog's automated classification reduces metadata inconsistency by standardizing terminology at publication time, improving data governance and discovery without significant workflow changes.