The Shift to Agentic Data Workflows

Google Cloud is positioning its data platform to support the "agentic era," where AI agents move beyond simple assistance to become proactive, autonomous participants in business decision-making. By 2027, Gartner predicts over 50% of business decisions will be augmented or automated by agentic AI. Google’s strategy focuses on three tiers: assistive tools for productivity, persona-specific agents for specialized tasks, and primitive APIs/SDKs for developers to build custom solutions.

Persona-Specific Data Agents

Google has introduced and matured several task-specific agents:

  • BigQuery Assistant: Integrated into BigQuery Studio, this tool assists with SQL authoring, troubleshooting, and performance optimization. It has driven a 30x increase in data processed by Gemini within BigQuery over the past year.
  • Data Engineering Agent: Now generally available, this agent allows engineers to build and troubleshoot complex data pipelines using natural language, while maintaining human oversight of the generated specifications.
  • Data Science Agent: Also GA, this agent operates within Colab notebooks to automate data exploration, feature engineering, and model validation, reducing tasks that previously took weeks to mere minutes.
  • Conversational Analytics: Available in Looker and BigQuery, this allows business users to query data, perform forecasting, and conduct anomaly detection using natural language, bypassing the need for schema knowledge.

Developer-Facing Tools and Extensibility

To address the need for custom agent development, Google has released several key tools:

  • Conversational Analytics API: Enables developers to embed conversational data capabilities directly into custom applications.
  • BigQuery SDK & MCP Server: The Model Context Protocol (MCP) server is a fully managed service that eliminates the need for custom glue code, allowing developers to connect BigQuery to open-source frameworks seamlessly.
  • Google Cloud Data Agent Kit: A new preview tool available as a VS Code extension and via CLI plugins (Gemini CLI, Cloud Code). It provides a unified interface for data practitioners to manage skills, explore data catalogs, and build pipelines without switching contexts.
  • Agent Analytics: A new plugin that allows developers to instrument their agents with a single line of code, persisting performance metrics (latency, token usage, sentiment) back into BigQuery for real-time monitoring.

Deep Data Research

Google introduced "Deep Dive," a feature within Gemini Enterprise that synthesizes structured data from BigQuery with unstructured and real-time web information. Unlike standard Q&A, this feature generates comprehensive, long-form reports and supports long-running agentic tasks.