BigQuery AI: The convergence of data and AI is here

From uncovering new insights in multimodal data to personalizing customer experiences, AI is emerging as the engine of modern innovation. The explosion in AI adoption has created a need to bring data and AI closer — not only to streamline the AI lifecycle, but also to bring AI-driven insights and workflow automation to everyone in the organization.

We created BigQuery ML to bring AI to your data, enabling data scientists and data analysts to build and deploy machine learning models directly inside BigQuery. Over the years, we built on this foundation by introducing capabilities such as AI-powered search, generative AI with SQL and many others. 

Today, we’re introducing BigQuery AI, which brings together BigQuery’s built-in ML capabilities, generative AI functions, vector search, intelligent agents, and agent tools. Using BigQuery AI, you can:

• Apply gen AI to your data : Bring Google and partner AI models directly to your multimodal data in BigQuery through simple SQL functions. 

• Simplify your data-to-ML journey: Manage your whole machine learning lifecycle in BigQuery — everything from feature engineering to model training, tuning, inferencing and monitoring, all without moving your data.

• Create workflows and apps faster: Whether you’re a data engineer, data scientist, or business user, you can accelerate your workflows with intuitive, role-specific agents built right into BigQuery.

Let’s take a closer look at the tools and technologies that fall under the BigQuery AI umbrella.

Unlock insights from multimodal data with generative AI

Bringing state-of-the-art AI models directly to your data through simple SQL commands can help you perform generative AI tasks as well as unlock deeper, semantic understanding from your multimodal data. 

• AI functions integrate LLMs and embedding models directly into your SQL queries, enabling you to perform tasks such as content generation, analysis, summarization, structured data extraction, classification, embedding generation, and data enrichment. You can also use AI functions for routine tasks such as filtering, rating, and classification. With managed AI functions, BigQuery chooses a model for you that is optimized for cost and quality. 

• Embeddings and search functions help you find information more intelligently. Traditional text search lets you quickly locate specific keywords in your data, but vector search allows you to search by meaning and context, not just exact words. This helps you uncover conceptually related items, finding relevant information that a simple keyword search would miss. Embeddings and vector search in BigQuery power use cases such as RAG, multimodal search, data deduplication, clustering and recommendation engines.

Data processing to AI inference all under one roof

When we first launched BigQuery ML, our goal was to bring AI and ML closer to your data, empowering SQL users to perform machine learning tasks directly in BigQuery, on their BigQuery data. Over the years, we added capabilities to provide a complete, end-to-end platform for accelerating the entire machine learning lifecycle.

Enterprises are using these capabilities to powerful effect. For instance, PUMA used BigQuery's integrated machine learning capabilities to advance beyond manual segmentation, crafting sophisticated audience segments based on purchase propensity. The outcome was hugely impactful: the top ML-derived audience segments demonstrated a remarkable 149.8% surge in click-through rate, a 4.6% uptick in conversion rate, and a 6% increase in average order value.

BigQuery AI streamlines the entire machine learning lifecycle by bringing the code to your data.

• No data movement: Train and run models directly in BigQuery using SQL or Python. No data movement or infrastructure management needed.

• End-to-end lifecycle: Handle everything from feature engineering to model training, evaluation, tuning , deployment and inference in BigQuery without needing expertise in specialized ML frameworks.

• Model flexibility: Choose from built-in models, import custom models you have trained in AI, or use pre-trained models (like TimesFM) for zero-shot inference.

• Unified inference: Seamlessly execute predictions via batch processing, real-time streaming, or remote inference.

And of course, BigQuery AI lets you use your preferred development environment — BigQuery Studio, the integrated AI-powered Colab Enterprise notebook, or an IDE of your choice.

Agentic experience for every data user

Under the BigQuery AI umbrella, we are also consolidating the data agents and assistive AI capabilities that are designed to streamline and automate workflows for various data professionals. 

• Data Engineering Agent allows you to build, modify, and manage data pipelines by describing your requirements in natural language. It translates your plain-language requests into production-ready SQL code, automating complex tasks like data cleaning, transformations, and schema modeling.

• Data Science Agent helps automate end-to-end data science workflows. It creates multi-step plans, generating and executing code, reasons about the results, and presents its findings. You also use it to generate visualizations with simple prompts, explain and transform code, as well as explain errors and fix them automatically.

• Conversational Analytics Agent empowers business users to bypass technical barriers, allowing them to ask questions in natural language and receive clear, actionable intelligence, truly democratizing data for all.

• Assistive AI features like data canvas and code completion simplify and speed up routine tasks. 

Beyond first-party agents and assistive AI capabilities, BigQuery also provides a powerful suite of tools for building custom agents and integrating agentic AI into your applications. The Conversational Analytics API provides the building blocks to embed natural language processing capabilities into your own products for tailored data experiences. And for more advanced use cases, the Agent Development Kit (ADK) offers a full-stack framework to build and deploy complex, multi-agent systems, while the Model Context Protocol (MCP) standardizes how AI models communicate with databases and other tools.

AI is changing how we all live and work, and nowhere is that more apparent than in how data professionals are approaching their jobs. BigQuery AI is a significant leap forward in how you can connect your data to AI. To learn more about BigQuery AI and get started with it, check out this guide. We can’t wait to see what you build next!