The role of data scientists is evolving from analyzing the past to building the future with intelligent, autonomous agents. This shift requires a move from analyst to agentic architect, but existing tools create friction and hinder creative flow. To address these challenges, new innovations are being introduced on an AI-native stack. These advancements aim to unify the development environment, enabling data scientists to move beyond analysis to action more efficiently.One key innovation is a single, intelligent notebook environment that integrates SQL, Python, and Spark, eliminating context-switching. Data scientists will now have native, SQL-based access to real-time and unstructured data essential for agent decision-making. The new tools facilitate moving from prototype to production in minutes, not weeks, with a complete toolkit for building, deploying, and connecting autonomous agents.Enhancements to Colab Enterprise notebooks in BigQuery and Vertex AI include native SQL and interactive visualization cells, transforming the notebook into an integrated development environment. The Data Science Agent has also been improved to incorporate sophisticated tool usage within its plans, making analysis more advanced and workloads more cost-effective. The Lightning Engine is now generally available, accelerating Spark performance significantly and integrating seamlessly with various tools.Stateful processing for BigQuery continuous queries will allow for complex, state-aware questions on live data, unlocking use cases like real-time fraud detection. Autonomous embedding generation in BigQuery over multimodal data simplifies building AI applications using vector databases. The "Build-Deploy-Connect" toolkit, including the Agent Development Kit (ADK), enables the creation of scalable, secure, and production-ready fleets of agents. Securely connecting these agents to enterprise data is streamlined through first-party BigQuery tools and the MCP Toolbox. The architect's workflow is also enhanced with Gemini CLI extensions for Data Cloud, allowing natural language interaction with data tasks directly in the terminal. These innovations empower organizations to automate tasks and become agentic architects, enabling them to sense, reason, and act with intelligence.