Introducing Dataset Q&A: Expanding natural language querying for structured datasets in Amazon Quick

Amazon Quick (QuickSight) has expanded its natural language querying with Dataset Q&A, bridging the gap between ad-hoc business questions and BI dashboards. The new capability lets any user ask questions in plain English against any structured dataset, without waiting for an analyst or pre-configured dashboard. The system translates the question into SQL, searches across all structured assets using a semantic graph to identify the correct source, and then generates a query that runs against the full dataset—no sampling, no data caps. It supports millions of rows, respects existing row-level and column-level security automatically, and returns aggregated results in seconds. Authors can also enrich datasets with business context via Dataset Enrichment, uploading field descriptions and custom instructions in YAML, JSON, or plain text, so the system understands ambiguous terms like “volume” correctly.

To build trust, the launch includes Chat Explainability, exposing the step-by-step reasoning behind each answer. Users can see the generated SQL, assumptions made, filters applied, and a plain-language explanation—removing the black box often associated with AI-powered BI tools. Dataset Q&A completes Quick's three-tier natural language suite alongside Dashboard Q&A (for published dashboards) and Topic Q&A (for curated field sets). The agentic system is purpose-built for enterprise complexity: it resolves lexical ambiguity, maps colloquial language to precise columns, and grounds queries against complex schemas without a predefined dictionary. For organizations struggling with BI backlogs, this directly empowers business users and frees analysts to focus on deeper work.