AWS Retools QuickSight's Q&A Layer, Pushing Business Context Down to the Dataset Level

Amazon QuickSight is changing where business meaning lives inside a BI deployment. AWS has introduced Dataset Enrichment, a feature that lets teams attach business context directly to individual datasets rather than bolting it on through a separate configuration layer, as the legacy Topics feature previously required.

With Dataset Enrichment, a dataset can now carry column descriptions, synonyms, calculated fields, custom instructions, and business rules natively. That means the semantic knowledge a BI team once had to define and maintain inside a standalone Topic (what a metric means, how a field should be labeled, which calculation to apply) now travels with the dataset itself. For any organization running multiple dashboards or Q&A experiences off the same underlying data, that consolidation removes a layer of duplicated setup and a common source of drift between what a dataset actually contains and what the Topic assumed it contained.

Topics Get a New Job

Rather than retiring Topics outright, AWS is repurposing the construct. Once enrichment is handled at the dataset level, Topics shift to acting as a semantic and reasoning layer that spans multiple datasets at once. Instead of being the place where individual field definitions and synonyms are configured, a Topic now becomes the coordination point for relationships, metrics, and shared business terminology that cut across several enriched datasets simultaneously.

Practically, this turns Topics into a single entry point for cross-dataset question answering. A user can pose a query in one conversation that draws on more than one enriched dataset, and QuickSight resolves it using the semantic context already embedded at the dataset level plus the relationship logic defined in the Topic. That is a meaningfully different architecture from the old model, where a Topic had to encapsulate everything, including field-level meaning, on its own.

Why the Split Matters

The separation of concerns is the real story here. Pushing column-level semantics down into the dataset means that context is defined once and reused everywhere that dataset appears, rather than redefined inside every Topic built on top of it. Topics, freed from that burden, can focus on the harder problem of reasoning across datasets: reconciling terminology, aligning metrics that might be named differently in different tables, and letting analysts ask questions that span data sources without manually joining or reconciling them first.

For teams currently running legacy Topics, this represents a migration path rather than a breaking change: business context that was scattered across Topic configurations can now be consolidated into the datasets themselves, with Topics reserved for genuinely cross-dataset reasoning. AWS frames this as a step toward semantic datasets, where meaning is a property of the data rather than a property of a downstream tool built to query it. That distinction matters for any team scaling natural-language BI beyond a handful of dashboards, since it is precisely the kind of duplicated configuration that becomes unmanageable at scale.