Conversational analytics in BigQuery reaches general availability
Google made conversational analytics in BigQuery and the Conversational Analytics API generally available, grounded in LookML, approved example queries, Knowledge Catalog and graph data.
Google grounds its data agent in four sources at once. Conversational analytics in BigQuery, and the Conversational Analytics API behind it, became generally available on July 28, 2026. Versions for AlloyDB, Cloud SQL and Spanner are in preview.
What shipped
- Plain-language questions over BigQuery, now generally available.
- Several grounding sources: LookML models, Golden Queries (approved example queries), Knowledge Catalog, and BigQuery or Spanner Graph.
- Works with Google's Agent Development Kit and with MCP, so the same capability can sit behind other agents.
Why it matters
Four sources means four places to record what the data means: a reporting model, a set of approved queries, a catalog and a graph. If the rule for a returned order lives partly in LookML, partly in a golden query and partly in a catalog entry, an agent can ground on all three and pick up three different rules. More grounding only helps when the sources agree. When they don't, the agent has no way to know which one is right.