Google enables Knowledge Catalog by default in Gemini

Google is enabling Knowledge Catalog by default in its Gemini application, starting 4 November, SiliconANGLE reported. The integration resolves business metric definitions and access controls before translating a user’s request into Structured Query Language (SQL), the language used to query a database.

Knowledge Catalog gives superintelligence (SI) agents a map of an organization’s business terms. Google says the agents apply the same governed formulas business teams use, such as those for net revenue or customer churn.

The update is part of Google’s Agentic Data Cloud, launched earlier in 2026 at Google Cloud Next to let agents access data where it resides rather than requiring it all in a central warehouse. The earlier cross-border Lakehouse also debuted there.

A separate preview for organizational context

Separate from the default integration, Google is previewing a self-learning enrichment feature for Knowledge Catalog. Sirish Chandrasekaran, Google’s vice president of product management for data analytics, says it continuously scans unstructured information in Google Workspace, Jira, SharePoint and GitHub and converts operational knowledge into Open Knowledge Format context sets for agents to use. He says Google expects autonomous agents to become the primary consumers of data instead of humans, motivating infrastructure that prioritizes low latency, zero data movement and automated governance.

SiliconANGLE reports that companies can also ground agents in semantic models for BigQuery, Snowflake, SAP, Looker and dbt.

Google says early adopters’ results show that this context can make the average agent four times as accurate while reducing consumption of tokens—the units used to process text—by up to 75%.

Google also describes reporting in Gemini that uses BigQuery and Knowledge Catalog to construct and save queries from business users’ requests. Chandrasekaran says teams can rerun saved reports on demand without token costs.

Data access and developer tools

The new borderless Lakehouse adds caching for data held on Amazon Web Services and Microsoft Azure. Chandrasekaran says companies can query that data over a secure private connection without data-transfer charges, while retrieved blocks are stored temporarily in Google Cloud to speed up later queries.

An AlloyDB integration also lets users query external tables in the Apache Iceberg format using standard PostgreSQL queries without copying the data. It connects live transaction data with historical lakehouse records.

Data Agent Kit is generally available, bringing Model Context Protocol tools into software development environments. Developers can describe a business goal in ordinary language and have agents create, refine and test data pipelines. Google says BigQuery Graph is also generally available, allowing petabyte-scale graph analytics within relational database tables without extract, transform and load (ETL) pipelines.

AlloyDB Express is in a broader preview, while the TimesFM 3.0 forecasting model and TabFM predictive analytics model are also in preview.