Amazon SageMaker Data Agent integrates business context into conversations
Ask "What data do I have on customer churn?" and get accurate SQL/Python code—no more deciphering cryptic table names.
View original announcement →Visual Summary
graph TD
A{{SageMaker Data Agent}}:::announced
B((Data Practitioner)):::external
C(SageMaker Catalog):::storage
D(AWS Glue Data Catalog):::storage
E(Amazon Redshift):::storage
F([Business Context Search]):::feature
G([SQL/Python Code Gen]):::feature
H([Data Governance]):::feature
I(SageMaker Unified Studio):::compute
J((Third-Party Catalogs)):::external
B ==>|"natural language"| I
I ==>|"routes query"| A
A -->|"searches technical"| D
A -->|"searches technical"| E
A ==>|"searches business terms"| C
J -.->|"syncs metadata"| C
A -->|"resolves tables"| F
F -->|"generates"| G
A -.->|"checks access"| H
classDef announced fill:#ff9900,stroke:#ec7211,color:#fff,font-weight:bold
classDef compute fill:#e3f2fd,stroke:#1565c0,color:#1565c0
classDef storage fill:#e8f5e9,stroke:#2e7d32,color:#2e7d32
classDef feature fill:#fff3e0,stroke:#e65100,color:#e65100
classDef external fill:#f5f5f5,stroke:#616161,color:#616161
What's New
Amazon SageMaker Data Agent now integrates with SageMaker Catalog business context and metadata, allowing data practitioners to discover datasets and generate SQL and Python code using natural business language instead of cryptic technical table names. The agent leverages glossary terms, custom metadata forms, asset summaries, and README content—including metadata synced from third-party catalogs like Collibra, Atlan, and Alation—to identify the correct tables and columns for any query. This integration also enforces data governance by checking subscription status and surfacing access request links when a user attempts to query restricted assets.
How It Works
- When a user submits a natural language question in a SageMaker Unified Studio notebook or Query Editor, the Data Agent performs a dual search: it queries technical metadata from AWS Glue Data Catalog and Amazon Redshift, and simultaneously queries SageMaker Catalog for business context matching the user's terminology.
- The agent merges technical and business metadata results to resolve the correct physical catalog, database, table, and column references that correspond to the business terms used in the question.
- Business context sources searched include glossary terms, custom metadata form values, asset summaries, and README content that organizations have curated in SageMaker Catalog—including assets synced from Collibra, Atlan, and Alation via configured metadata sync integrations.
- Once the correct tables are identified, the agent generates accurate SQL or PySpark code on the first attempt, using fully qualified references and planning multi-step workflows with the correct sequence of tables and transformations.
- For assets the user cannot access, the agent respects data governance policies by checking subscription status and providing direct access request links rather than generating code against unauthorized data.
- The integration is surfaced natively within SageMaker Unified Studio notebooks and the built-in Query Editor, requiring no changes to existing data workflows or catalog curation processes.
Why It's Important
- Organizations that have invested significant time curating business metadata in SageMaker Catalog—or syncing it from enterprise catalogs like Collibra, Atlan, or Alation—can now directly leverage that investment inside their coding and analytics workflows, maximizing ROI on catalog initiatives.
- Data practitioners no longer need to know cryptic internal table names or schema structures; they can express intent in business language (e.g., "Calculate customer retention rate"), dramatically reducing the time spent on data discovery before any analysis begins.
- First-attempt code accuracy improves because the agent grounds code generation in verified business context rather than relying solely on table name inference, reducing costly iteration cycles and debugging time.
- Governance is enforced at the point of code generation, not as an afterthought—users are informed of access restrictions and given a path to request access, reducing accidental exposure of sensitive data.
- The capability bridges the gap between business stakeholders and technical data teams by enabling a shared vocabulary, making it easier for less technical users to participate in data workflows.
How It's Different
- Unlike generic AI coding assistants or text-to-SQL tools that rely solely on schema introspection, SageMaker Data Agent enriches code generation with curated business metadata (glossary terms, summaries, READMEs), producing semantically correct code rather than syntactically plausible but contextually wrong code.
- The integration is catalog-aware and governance-aware simultaneously, combining data discovery, code generation, and access control in a single conversational interaction—a capability not typically bundled together in standalone AI coding tools.
- Native support for third-party catalog metadata sync (Collibra, Atlan, Alation) means enterprises are not locked into AWS-only metadata workflows; existing enterprise catalog investments are honored and reused rather than replaced.
- The agent plans multi-step analytical workflows—not just single-table queries—understanding the correct sequence of joins and transformations needed to answer complex business questions, which goes beyond simple text-to-SQL capabilities.
- Because the integration lives inside SageMaker Unified Studio, it works across both SQL (Query Editor) and Python/PySpark (notebooks) surfaces, providing a consistent natural-language experience regardless of the compute or query engine being used.
When to Prefer It
- Use this capability when your organization has already invested in curating a SageMaker Catalog with business metadata, glossary terms, or data product definitions, and you want that investment to accelerate day-to-day analytics work.
- Ideal for data teams onboarding new analysts or data scientists who are unfamiliar with the internal naming conventions of the data warehouse or lakehouse, allowing them to be productive immediately using business language.
- Best suited for organizations running enterprise data catalogs (Collibra, Atlan, or Alation) that want a seamless bridge between their governance and metadata layer and their active coding and analytics environment.
- Choose this integration when data governance and access control are critical requirements, as the agent enforces subscription-based access policies at query time rather than relying on users to self-police.
- Particularly valuable for complex analytical use cases requiring multi-table joins and sequential transformations, where manually identifying the right tables and their relationships would otherwise require significant tribal knowledge.
- Use it when reducing time-to-insight is a priority—eliminating the discovery phase of analytics (finding the right tables) directly compresses the overall time from business question to actionable result.
Availability
- Status: Generally Available (GA) as of June 4, 2026.
- Supported surfaces: Available in SageMaker Unified Studio notebooks and the built-in Query Editor.
- Supported regions: US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (Ireland), Europe (Frankfurt), Europe (London), Europe (Paris), Europe (Stockholm), Asia Pacific (Tokyo), Asia Pacific (Seoul), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Mumbai), South America (São Paulo), and Canada (Central)—all regions where Amazon SageMaker Unified Studio is currently available.
- Prerequisites: Requires an Amazon SageMaker Unified Studio domain (IAM or IAM Identity Center), a SageMaker Catalog configured with published assets and data products, and assets enriched with business metadata (glossary terms, metadata forms, summaries, or READMEs).
- Third-party catalog support: Collibra, Atlan, and Alation are supported via configured metadata sync integrations with SageMaker Catalog.
- Pricing: No separate pricing information was announced; costs are expected to follow existing SageMaker Unified Studio and SageMaker Data Agent pricing models.
- Limitations: Business context search is only as effective as the quality and completeness of metadata curated in SageMaker Catalog; sparse or uncurated catalogs will limit the agent's ability to resolve business terms to correct tables.
