Google adds agent tools to BigQuery, AlloyDB and Gemini data

Google Cloud announced updates to its Agentic Data Cloud, a set of data tools for AI agents. Knowledge Catalog is on by default in Gemini, in preview. It finds trusted tables and business terms before Gemini writes a database query. Google also announced AlloyDB Express and PostgreSQL for agents in AlloyDB, the two in preview. BigQuery Graph is available to all users.

What is new

Knowledge Catalog reads definitions such as net revenue from BigQuery, Looker, Databricks, Snowflake, dbt and SAP, and does not move the data. It also reaches catalogs from Oracle, Snowflake, Databricks, Atlan and Collibra. Gemini can write PySpark code and train models. Two models from Google Research, TimesFM 3.0 and TabFM, operate in BigQuery in preview.

The numbers

These are claims by Google. Google gave no test method.

• PostgreSQL for agents: more than 3 million queries each second.
• BigQuery: queries more than 40% faster and more than 50% cheaper in one year.
• Managed Spark: value for cost more than 3x better than other tools.

What is not known

Google gave no precision results for Knowledge Catalog. It does not say what the other catalogs must do to work with it. Some parts are only in preview.

Insight: Google wants to define the data meaning for agents

Google reaches into the data of rivals: Oracle, Snowflake and Databricks. It also adds Looker metric formulas to the catalog, and it works with Oracle on a connection for agents. This reads as a move to own the layer that tells what data means, and each agent then asks Google first. The one that sets the definitions gets control of the agents that use them.