Amazon Quickに「Agentic Catalog Experience」が登場Announcing the Agentic Catalog Experience in Amazon Quick

Amazon has introduced an "Agentic Catalog Experience" for Amazon Quick, an AI-powered workflow that lets data curators discover upstream catalog assets through natural language and automatically generate analytics-ready content. Announced on the AWS Machine Learning Blog and currently available in preview, the feature is notable because it applies agentic AI to data preparation, one of the more manual and repetitive stages of building business intelligence.

The workflow addresses a common bottleneck. Before analysts can explore data or build dashboards, someone typically has to locate the right source tables, understand their structure, and shape them into reusable objects. The Agentic Catalog Experience aims to compress that process: a data curator can describe the data they need in plain language, and the system searches connected catalogs for matching upstream assets. From there, it can auto-create Datasets and Topics, the building blocks Amazon Quick uses for analysis and natural-language querying.

A key detail is the notion of inherited semantics. Rather than starting from a blank slate, the generated Datasets and Topics carry over metadata and semantic context from the source catalog. That likely includes information such as table and column descriptions, relationships, and business-friendly naming that already exist in the upstream system. Preserving this context is intended to reduce duplicated effort and help ensure that downstream analytics reflect the definitions maintained by data governance teams.

At launch, the preview supports two catalog systems: AWS Glue Data Catalog and Databricks Unity Catalog. AWS Glue Data Catalog is Amazon's managed metadata repository, widely used to register tables for services such as Amazon Athena, Amazon Redshift Spectrum, and Amazon EMR. Databricks Unity Catalog is the governance and metadata layer within the Databricks platform, providing centralized access control and lineage across data and AI assets. Supporting both suggests Amazon is positioning the feature to work across mixed environments rather than only AWS-native data stores.

The word "agentic" points to the broader direction of the tooling. Instead of a single natural-language query returning a chart, an agentic workflow can carry out multiple steps on the user's behalf, such as searching, interpreting, and creating objects. While AWS has not detailed the underlying models in this summary, the company's generative AI features are commonly built on Amazon Bedrock, its managed service for foundation models, and the tags associated with the announcement reference Bedrock alongside data-catalog and natural-language capabilities.

Amazon Quick sits within Amazon's analytics and business intelligence portfolio, and Topics in particular connect to natural-language Q&A, where users ask questions in everyday language and receive answers drawn from a defined semantic model. By automating the creation of these Topics from cataloged data, the new experience appears designed to shorten the path from raw governed data to self-service analytics, lowering the amount of manual modeling a curator must perform.

The move also fits a wider industry pattern. Vendors across the data and BI space have been embedding AI assistants and agents into catalog, preparation, and dashboarding tools, framing natural language as the primary interface for both technical and non-technical users. Data catalogs themselves have grown in importance as organizations try to manage sprawling data estates, enforce governance, and track lineage, making them a logical