What’s new in the Agentic Data Cloud: Powering the System of Action

Companies are shifting from gen AI that simply answers questions to autonomous agents that perceive, reason, and act on their behalf. Attempting to scale these agents on legacy stacks exposes structural failures that can lead to fractured governance, a persistent trust gap, and broken reasoning loops, all while causing costs to spiral.

To solve this, we’re introducing the Agentic Data Cloud: an AI-native architecture that evolves the enterprise data platform from a static repository into a dynamic reasoning engine. It closes the gap between thinking and doing, allowing AI agents to act on your business data and context. While last-generation systems of intelligence were built only for human scale, the Agentic Data Cloud is a System of Action, built for agent scale.

Leading organizations are already using the Agentic Data Cloud to deliver tangible value:

• **Vodafone**has launched hundreds of agents to deliver uninterrupted service to their customers, which is expected to save them millions of euros every year.

• **American Express** is moving a core on-premises data warehouse and hundreds of production applications to BigQuery to power trusted agentic commerce at scale.

• **Virgin Voyages** is using over 1,000 specialized AI agents, including one that slashes mass itinerary rebooking from six hours to just 11 minutes.

Today, we’re announcing three new innovation areas powering our Agentic Data Cloud:

• **A universal context engine** that provides agents with trusted business context to drive higher accuracy.

• **Agentic-first practitioner experiences** to evolve the role of data practitioners and developers as orchestrators of agents.

• **An AI-native, cross-cloud lakehouse** that eliminates data silos by connecting your entire data estate.

**Enabling agents with a universal context engine**

AI is only as smart as its context. If an agent doesn't understand your definition of "margin" or the intricate relationships in your supply chain, it’s forced to guess. In the age of the agentic enterprise, data alone is not enough, and the old model of governance is insufficient. This is why we evolved the Dataplex Universal Catalog into the Knowledge Catalog, which maps and infers business meaning across your entire data estate, using a rigorous framework of aggregation, continuous enrichment, and search.

Here’s how it works:

• **Aggregation:**To build true context, you must bring it together from everywhere. We are aggregating native context across your Google Cloud and partner data platforms. This includes third-party catalogs, applications, operating systems, and AI platforms like Palantir, Salesforce Data360, SAP, ServiceNow, and Workday (Preview). By using our Lakehouse, your third-party data assets are automatically mapped to the Knowledge Catalog. For Google Cloud sources, we are automating business logic with the new LookML Agent (Preview), which autonomously generates semantics from strategy docs, and BigQuery measures (Preview), which embeds that business logic natively into the platform.

• **Continuous enrichment**: The Knowledge Catalog delivers continuous enrichment by analyzing usage logs across your organization and profiling data behind the scenes. It learns how your enterprise actually uses data, not just what it is. This extends to unstructured data. The moment a file lands in Google Cloud Storage, our Smart Storage (Preview) instantly tags and enriches images and, soon, PDF objects. The Knowledge Catalog also identifies useful collections of unstructured data and uses Gemini to automatically generate missing schemas, mapping complex relationships so your AI is no longer flying blind.

• **Search and retrieval:**Creating a massive context layer is great, but in the agentic era, search has evolved to be the new query path. The hardest problems at enterprise scale are speed, relevance, global reach, and security. To solve this, the Knowledge Catalog uses a sophisticated hybrid search stack built on Google Search innovation. To deliver relevance, it combines semantic and lexical matching with intelligent, machine-learning-based re-ranking. To deliver trust, we are enforcing your security permissions natively with access-control-aware search, so agents can only retrieve and act on the assets they are authorized to see. This high-precision infrastructure instantly identifies trusted context and feeds it to specialized agents.

The Knowledge Catalog now powers the Deep Research Agent (Preview). Part of the suite of Google-made agents available in Gemini Enterprise, this agent can perform multi-step reasoning across Google Cloud data platforms, such as BigQuery, as well as internal documents and web assets to answer complex questions with citations and precision that previously required weeks of manual effort.

**Agentic-first practitioner experiences**

As we move to this new architecture, the data practitioner role shifts from writing manual pipelines to orchestrating intent-driven engineering.

We’re accelerating this transition with the **Google Cloud Data Agent Kit** (Preview). Rather than introducing a new interface, we are launching a portable suite of skills, tools, environment-specific extensions, and built-in plugins, that drop into the environments developers love. By meeting practitioners where they already build — including VS Code, Gemini CLI, Codex, and Claude Code — we turn your IDE, notebook, or agentic terminal into a native data environment. This enables your environment to autonomously orchestrate a wide range of business outcomes, automatically selecting the right frameworks (e.g., dbt, Apache Spark, or Apache Airflow) and generating production-ready code based on Google’s gold standards.

This kit doesn't just connect tools — it injects high-performance capabilities directly into the developer's flow, scaling to petabytes without moving data. In fact, the Data Agent Kit features the same skills and tools that power our own out-of-the-box agents, including:

• **Data Engineering Agent** (GA): Builds complex pipeline transformations from scratch and enforces governance rules to keep bad data out of production.

• **Data Science Agent** (GA): Automates the model lifecycle — from wrangling to training — scaling across BigQuery Dataframes and Serverless Apache Spark.

• **Database****Observability Agent** (Preview): Acts as a 24/7 guardian for your infrastructure, diagnosing root causes and executing database remediations.

To help ensure the smooth execution of agents, Google Cloud has fully embraced Model Context Protocol (MCP), which provides a secure, universal interface that allows any agent to discover and use your data assets across our core engines, including: BigQuery**,**Spanner(Preview)**,**AlloyDB,Cloud SQL(GA),and Looker MCP (Preview)**.** MCP for Google Cloud uses our security stack, governing agent interactions based on your existing IAM policies, VPC Service Controls, and data residency requirements.

We’re also reimagining the business user experience with Conversational Analytics, now supported across BigQuery (GA), Cloud SQL, Spanner, AlloyDB (Preview), and Looker (GA). Organizations can simply publish these custom analytical agents in Gemini Enterprise, enabling employees to chat with live data in a familiar interface. By removing the technical barriers, we’ve eliminated the weeks spent waiting for manual reports, allowing businesses to move at the speed of thought.

**A cross-cloud foundation built for agentic scale**

For an agent to act, it must have a fundamentally open foundation. If an agent is blocked by cross-cloud latency or trapped in a proprietary walled garden, its autonomy is broken. That’s why we are introducing a trulyborderless, cross-cloud Lakehousethat liberates your data wherever it resides by:

• **Connecting analytical estates**: We’re integrating Cross-Cloud Interconnect (CCI) directly into our data plane. By combining CCI’s dedicated, high-speed private networking with Apache Iceberg REST Catalog**,** we’re enabling connectivity across clouds that is low latency and eliminates massive egress fees. As a result, agents can use data across AWS and Azure as if it were local to Google Cloud with seamless cross-cloud access.
• **Ending proprietary silos**: We’re championing open federation to end the era of proprietary catalogs by launching bi-directional federation (Preview). Powered by the Iceberg REST Catalog, engines can now read directly from Databricks Unity Catalog on Amazon S3 (Preview), Snowflake Polaris (Preview), and the AWS Glue Data Catalog on Amazon S3 (Preview)**.** This is reinforced by enhanced Lakehouse Governance (Preview), which ensures your security policies and access controls apply instantly across this borderless environment.
• **Unlocking operational data**: We’re announcing Spanner Omni (Preview), unchaining the most scalable, globally consistent database on the planet. For the first time, you can run the Spanner engine anywhere — across clouds, on-premises, or on your laptop — with the same capabilities used to run Google.
• **Bridging the insight to action gap:**We’re also closing the gap between insight and action. Most "unified" data platforms force the creation of complex ETL pipelines that block agents from accessing your real-time data. With Lakehouse federation for AlloyDB(Preview), we’re removing these pipelines by providing protocol-level, zero-ETL synchronization to give agents access to deep analytical history with low latency in operational transactions.

Automating the future

Moving to agent scale generates orders of magnitude more workloads. To support this, we are announcing four major performance breakthroughs:

1. **Lightning Engine for Apache Spark**delivers up to 2x the price-performance over the proprietary market alternative.

2. **Managed Lustre** delivers up to 10 terabytes-per-second of throughput to make sure data moves quickly enough for demanding models.

3. **Bigtable**now supports an in-memory tier that delivers sub-millisecond read latency for real-time applications. This means you can finally eliminate separate, side-by-side caching layers.

4. **BigQuery fluid scaling** helps lower costs by up to 34% on average for autoscaling workloads, scaling up resources instantly when agents act, and scaling back when they don't.

Build your success on a System of Action

The era of passive observation is over. The future belongs to the System of Action, made possible by Google’s Agentic Data Cloud.

**Ready to see what an Agentic Data Cloud can do for your business?**

Sign up for a strategy workshop today on how to get your data ready to fuel autonomous agents through a System of Action.

**Want to learn more about an Agentic Data Cloud?**Read our blueprint for turning passive data into proactive action.

Posted in

• Data Analytics
• Databases
• Business Intelligence
• Google Cloud Next