Agent-ready analytics: Unlocking insights with BigQuery augmented analytics

BigQuery now includes six new Table-Valued Functions (TVFs) that automate complex data analysis by combining AI, machine learning and statistical methods. These functions operate where the data resides, eliminating the need to export datasets to external tools. They generate structured SQL outputs, enabling integration into automated workflows for AI agents. The functions address specific analytical challenges such as identifying metric drivers, quantifying event impacts and detecting structural shifts in time-series data.

The new functions include ML.DETECT_CHANGE_POINTS, which identifies statistically significant shifts in time-series data, and AI.KEY_DRIVERS, which pinpoints the top factors behind metric changes. AI.CAUSAL_EFFECT measures the impact of actions by comparing observed results to expected baselines. Other functions analyze correlations, isolate underlying trends and detect predictable cycles. These tools can be chained together to perform multi-step analyses without manual intervention.

A practical example demonstrates how these functions work together using the Austin Bikeshare dataset. ML.DETECT_CHANGE_POINTS identified a structural shift in February 2018, coinciding with the Austin City Council’s ‘Dockless Mobility Pilot Program’. The shift aligned with the integration of shared electric scooters and bikes into the city’s transit system. This change point was then used as input for AI.KEY_DRIVERS to determine the factors driving the surge in bikeshare trips.

The functions are designed to scale across millions of individual time series, making them suitable for large datasets. They support automated, conversational data investigation workflows by producing structured outputs that AI agents can interpret. The integration with BigQuery ensures faster analysis by reducing data movement and enabling real-time insights.