The AI gold rush is moving from model training to data wrangling, and Google is betting that SQL is the pickaxe of choice. With the introduction of three new managed AI functions within BigQuery, Google aims to democratize AI-driven data analysis, letting everyday analysts wield the power of large language models (LLMs) directly within their existing SQL workflows
Forget exporting data to data scientists and waiting for the magic to happen. These new functions promise to collapse complex workflows into single queries, turning any analyst into an AI-powered insights engine.
The core of Google’s announcement lies in the simplicity these functions offer. Dubbed AI.IF, AI.CLASSIFY, and AI.SCORE, these tools are designed to let users filter, categorize, and rank data based on semantic meaning, all without the headache of prompt engineering or juggling external tools.
Imagine using AI.IF in a WHERE clause to filter customer reviews based on sentiment, or employing AI.CLASSIFY to automatically categorize support tickets. The goal? To make AI accessible to the SQL-fluent masses.
Traditionally, extracting insights from unstructured data required a complex dance. Data analysts would often have to hand off raw data to data scientists, who would then massage it with machine learning models before returning it for SQL-based analysis. According to Bradley Shimmin , lead of the data, analytics, and infrastructure practice at The Futurum Group, these new functions “can literally collapse that entire workflow into a single query, using standard SQL syntax.”
The bottleneck isn’t just time; it’s also talent. Not every organization has a team of data scientists ready to tackle every analytical challenge. These functions aim to empower existing analysts to do more with less specialized expertise.
Managed AI: Google Takes the Wheel
A key selling point is the “managed” nature of these functions. As HyperFRAME Research’s practice leader of AI stack, Stephanie Walter, points out, Google handles the backend complexities: model selection, prompt optimization, query plan tuning, and endpoint management.
This abstraction is crucial. Instead of wrestling with the intricacies of AI model deployment, analysts can focus on what they do best: extracting insights from data.
“This managed approach addresses the enterprise pain-point of complexity and operational risk: instead of analysts or teams having to decide which model variant to use, and optimize queries for latency and cost, Google abstracts that,” Walter said.
Google isn’t alone in this pursuit. The integration of AI into data warehouses is a growing trend, with major vendors vying for dominance. Databricks already offers AI Functions accessible from SQL or Python. Snowflake provides AI_PARSE_DOCUMENT, AISQL , and
