Google Cloud Brings BigQuery Conversational Analytics to General Availability

What Changed

BigQuery Conversational Analytics lets users chat with data agents using natural language to ask questions about BigQuery data. The Conversational Analytics API is accessed through geminidataanalytics.googleapis.com or regional endpoints. Developers use it to build AI-powered chat interfaces and data agents on top of that same capability.

Both were previously in preview. General availability removes that restriction for BigQuery.

What the Feature Does

The core mechanism is a data agent. A team creates an agent and connects it to knowledge sources such as tables, views, graphs, or user-defined functions in BigQuery. The agent can be configured with custom metadata, business term definitions, synonyms, and verified queries that encode how the organization's data should be interpreted.

Users then ask questions in plain language. The agent generates SQL, returns answers as text or charts, and shows its reasoning. Multi-part questions that use business terms work without requiring the user to specify field names or filter conditions directly.

The API also supports data sources beyond BigQuery. Through its QueryData method, it can connect to Looker, Data Studio, AlloyDB, GoogleSQL for Spanner, Cloud SQL, and Cloud SQL for PostgreSQL.

Within agent responses and verified queries, the feature supports AI and ML functions including forecasting, anomaly detection, and key driver analysis. Semantic search and semantic scoring are also supported.

Scope and Status

General availability applies specifically to BigQuery Conversational Analytics and the Conversational Analytics API. Conversational Analytics in Data Studio is available in a new experience to all Data Studio users. The Code Interpreter within that experience is separate: it requires a Data Studio Pro subscription with Gemini in Data Studio enabled. Conversational Analytics in Databases remains in preview.

Gemini in Looker, which supports natural-language data questions inside Looker, was already generally available before this announcement.