Google's agentic data push

Google used Cloud Next on April 22 to pitch a new “Agentic Data Cloud,” a data stack built for software agents to take actions, not just answer questions. (cloud.google.com) In plain terms, Google is trying to turn enterprise data systems from filing cabinets into control rooms: agents pull context from company data, decide what to do, and trigger workflows across apps and databases. Google said the new stack is built around three pieces: a “universal context engine,” agent-focused tools for practitioners, and a cross-cloud lakehouse. (cloud.google.com) The context layer is the renamed Knowledge Catalog, formerly Dataplex Universal Catalog, which Google says can aggregate metadata from Google Cloud and outside systems including Palantir, Salesforce Data360, SAP, ServiceNow and Workday in preview. Google announced the rename on April 10 and describes the product as an “AI-powered context graph” rather than a passive metadata registry. (cloud.google.com) (docs.cloud.google.com 1) (docs.cloud.google.com 2) The data layer underneath is BigQuery and Google’s lakehouse tooling, which now leans harder on Apache Iceberg so the same data can be used across engines and clouds without constant copying. Google’s April 22 updates included managed Iceberg tables, multi-table transactions, change data capture and cross-cloud features tied to BigQuery Omni. (cloud.google.com) (docs.cloud.google.com) For agents, speed matters because they are supposed to react while data is still fresh. BigQuery’s continuous queries run SQL continuously on incoming data and can send results to Pub/Sub, Bigtable, Spanner or BigQuery tables, giving agents a way to watch events and trigger next steps in near real time. (docs.cloud.google.com 1) (docs.cloud.google.com 2) Google is also pushing a common language for agents to talk to each other. It introduced the open Agent2Agent, or A2A, protocol on April 9, 2025 with more than 50 partners, and by July 31, 2025 Google said support had grown to more than 150 organizations. (developers.googleblog.com) (cloud.google.com) A2A is meant to solve a practical problem: companies are building agents in different tools, and those agents need a standard way to discover each other, exchange information and coordinate tasks securely. Google positions A2A as complementary to Anthropic’s Model Context Protocol, which focuses on giving agents access to tools and context. (developers.googleblog.com) (cloud.google.com) Google tied the pitch to customer examples rather than only demos. It said Vodafone has launched hundreds of agents, Virgin Voyages is using more than 1,000 specialized agents, and American Express is moving a core on-premises warehouse and hundreds of production applications to BigQuery. (cloud.google.com) The company has been laying the groundwork for a year. At Cloud Next 2025, Google introduced A2A and its Agent Development Kit, said more than four million developers were building with Gemini, and reported Vertex AI usage had increased 20-fold over the prior year. (blog.google) The bet now is that enterprise data platforms will be judged less by how well they answer dashboards and more by how safely they let agents do work. Google’s answer is more orchestration, more shared context and more controls wrapped around the same data stack. (cloud.google.com) (venturebeat.com)