Google Embeds AI Agents into Core Data & Dev Platforms

The landscape of enterprise technology is undergoing a profound transformation, driven by the emergence of “agentic AI.” Far from being mere theoretical concepts or experimental prototypes, these intelligent AI agents are increasingly assuming tangible operational roles within production environments. From orchestrating complex data pipelines to generating code and interpreting intricate business logic, these autonomous systems are redefining workflows, prompting tech giants like Google to rapidly establish the foundational infrastructure for this burgeoning field.

Google’s latest foray into this space was unveiled at its Cloud Next Tokyo event, where the company introduced a comprehensive suite of AI agents and significant infrastructure enhancements. These new capabilities are designed to empower data engineers, scientists, analysts, and developers alike. Rather than presenting standalone tools or entirely new interfaces, Google has seamlessly integrated these preview releases directly into its core platforms, including BigQuery, Vertex AI, and GitHub. This strategic integration is bolstered by updates that embed vector search and large language model reasoning directly into Google’s data services, ensuring immediate utility without disrupting existing workflows. While some of these new AI agents operate discreetly in the background, others engage more directly, yet all share a singular objective: to drastically reduce the time spent on repetitive tasks.

In the realm of software development, the Gemini CLI stands out, offering robust support for teams working within GitHub. It streamlines critical processes such as pull request reviews, issue triage, and handling minor coding tasks channeled through comments. A simple mention in an issue is all it takes for Gemini CLI to return with proposed code, accompanying tests, and a draft change ready for review. Complementing this agent, Google is also releasing open-source workflows designed to automate common chores like labeling or sorting incoming issues. This seemingly minor shift promises to significantly prevent teams from becoming bogged down by accumulating backlogs.

For data engineers, Google has introduced the Data Engineering Agent within BigQuery, a tool engineered to eliminate manual pipeline setup. Instead of crafting SQL queries from scratch or navigating between multiple interfaces, users can simply describe their requirements in plain language—for instance, “Load a CSV, clean specific columns, and join it with another table.” The agent then autonomously manages the entire workflow from initiation to completion. Google emphasizes that this innovation is not intended to displace engineers but to accelerate the most repetitive and time-consuming aspects of their work. The agent’s output remains fully editable, ensuring engineers retain complete control, thereby enabling teams to move faster, particularly when dealing with large volumes of complex or fragmented data.

This principle of acceleration extends seamlessly to data science with the introduction of a new Data Science Agent, integrated into Colab Enterprise. Designed to support the full lifecycle of exploratory analysis and modeling, this agent connects directly with BigQuery and Vertex AI. It responds to natural language prompts for diverse tasks, including data profiling, feature generation, and the execution of machine learning models. What truly distinguishes this agent is its ability to follow through on each step as part of a continuous workflow. Google asserts that it can plan, execute, reason, and present findings within a single session, allowing teams to review, refine, and guide the results without losing momentum. As with its other agents, Google reiterates that the goal is not to replace data scientists but to significantly expedite their process during the initial, often repetitive, stages of experimentation.

For business analysts, Google is enhancing its Conversational Analytics Agent with a new Code Interpreter. This powerful tool translates natural language prompts into executable Python code, performs the requested analysis, and then presents both the results and corresponding visualizations. It is specifically designed to tackle complex questions that extend beyond simple SQL queries, such as customer segmentation or forecasting. Google’s aim here is to empower teams to transition from vague questions to structured insights without the need to write or manage code themselves.

Underpinning these new agent capabilities are substantial infrastructure upgrades across Google’s data stack. Spanner now features a columnar engine optimized for analytical workloads, delivering performance gains of up to 200 times on certain queries. BigQuery is gaining improved access to live transactional data through a new feature called Data Boost. Furthermore, Google is directly embedding vector search and retrieval-augmented generation (RAG) into its platform, providing agents with persistent memory that remains firmly grounded in real company data.

“The way we interact with data is undergoing a fundamental transformation, moving beyond human-led analysis to a collaborative partnership with intelligent agents,” stated Yasmeen Ahmad, Managing Director for Google Cloud’s Data business, in a blog post announcing the launch. She described this as the “agentic shift,” heralding a new era where specialized AI agents operate autonomously and cooperatively to unlock insights at previously unimaginable scales and speeds.

Beyond the agents themselves, Google is actively laying the groundwork for broader adoption. The new Gemini Data Agents API, launching initially as the Conversational Analytics API, will enable developers to embed Google’s agentic capabilities directly into their own tools and workflows. Alongside this, the Data Agents API and Agent Development Kit (ADK) will empower teams to construct custom agents from the ground up, tailored to their unique internal logic and business requirements. These tools transform the agentic model from something teams simply use into something they can actively shape. To ensure these agents operate within defined boundaries, Google is also rolling out the Model Context Protocol (MCP) and the Looker MCP Server, which guarantee that agents working with structured data adhere to the correct context, permissions, and definitions.

The introduction of these agents and supporting tools underscores Google’s fundamental rethinking of how people interact with data. The overarching objective is not to replace existing workflows but to make them inherently faster, lighter, and more focused. This strategic shift promises to significantly reduce the time teams spend on setup and mundane tasks, allowing them to dedicate more energy to solving genuine business problems.