Bringing conversational analytics to Gemini with Agentic Data Cloud connectors

Google Cloud has launched Data Cloud connectors in Gemini for business, currently in preview, allowing users to query live enterprise data using natural language directly from their workspace. The connectors enforce row- and column-level security and respect end-user credentials, addressing limitations of earlier generative AI tools that struggled with complex schemas or governance rules. By integrating with Knowledge Catalog, the system interprets user prompts, locates relevant data, and executes permission-scoped queries under the user's identity, streamlining data access without requiring technical expertise.

The connectors support querying across multiple data sources, including BigQuery, operational databases like Spanner, AlloyDB, and Cloud SQL, and governed data from Looker. Users can pull data directly into everyday tools such as Google Docs, Gmail, or Google Slides for further analysis or reporting. The system uses semantic discovery and context mapping to translate plain-language questions into precise database queries, ensuring accuracy while reducing latency and token consumption through curated business context from Knowledge Catalog.

Beyond native storage, the BigQuery connector unifies data across environments, including open lakehouse tables in cloud object storage and cross-cloud datasets in AWS or Azure. It inherits fine-grained access policies from BigQuery and Knowledge Catalog, applying row- and column-level controls whether querying native tables or open lakehouse files. This approach eliminates physical data silos, enabling a single conversational interface to access diverse datasets while maintaining governance and security standards.

Administrators can configure Data Cloud connectors in the Gemini admin console, setting up default read-only connections to datasets in BigQuery, operational databases, or Looker. End users then enable the connector in Gemini to start querying data and building reports in plain language. Unlike data ingestion modes, these connectors use federated query mode, querying data in place without copying or indexing, while leveraging Knowledge Catalog to search accessible metadata and business context automatically.