BigQuery has introduced a suite of augmented analytics Table-Valued Functions (TVFs) that automate insight discovery and pattern explanation using AI, ML, and statistical methods. Because the functions run directly where data lives in BigQuery, they speed up analysis and reduce the need to export data into external tools. Their compact structured SQL output also makes them easy to integrate as skills for AI agents, enabling automated, conversational data investigation workflows.

The six new functions address distinct analytical challenges. AI.KEY_DRIVERS identifies the top drivers behind an increase or drop in a metric between two time periods or groups, such as why revenue spiked in a quarter. AI.CAUSAL_EFFECT quantifies the impact of an action or event by comparing observed results with an expected baseline. ML.CORRELATION evaluates the direction and strength of relationships between numeric metrics. ML.DETECT_CHANGE_POINTS identifies specific dates or intervals where a metric experiences a structural shift. ML.TREND separates underlying growth or decline from short-term noise. ML.SEASONALITY discovers repeating cycles across hours, days, weeks, months, or quarters.

The article demonstrates chaining these functions using the Austin Bikeshare sample dataset. Step 1, ML.DETECT_CHANGE_POINTS detects statistically significant shifts in the daily baseline; it is highly scalable and capable of running across millions of individual time series. The output highlights two change points with the longest durations, including one in February 2018 that aligns with the Austin City Council’s Dockless Mobility Pilot Program and an Austin Bikeshare promotional partnership with the University of Texas offering free annual memberships to UT students. Step 2, AI.KEY_DRIVERS scans millions of rows to attribute the surge, defining the post-shift slice as the interest group and the pre-shift period as the reference group. The analysis shows overall trip count increased +374.7% (+40,159 trips), with growth overwhelmingly concentrated in U.T. Student memberships (+7,167.1%) and trips ending at the 21st & Speedway @PCL station (+20,739.1%). Step 3, AI.CAUSAL_EFFECT constructs an ARIMA_PLUS counterfactual to measure what trip volume would have been without the program, isolating true return on investment over organic expectations. It reveals the program caused a +358% volume surge above the organic baseline, an estimated 89,775 incremental trips, with a 99.9% probability of causal effect.

All new augmented analytics functions are available in BigQuery‘s Conversational Analytics, allowing users to chat with agents about their data in natural language. Because the TVFs execute complex analytics at BigQuery scale in seconds, Conversational Analytics can orchestrate multi-step investigative workflows from a single prompt. One example on Chicago taxi trips uses ML.CORRELATION with AI.KEY_DRIVERS. The results show credit card payments are the primary positive driver of trip distance, adding +1.65M miles, with trips from O’Hare International Airport contributing another +1.10M miles among tipped credit card rides; cash transactions act as a significant negative driver (-652.96K miles) for shorter journeys. Another example on Iowa liquor sales uses ML.TREND with ML.SEASONALITY, revealing persistent long-term growth from 1.3–1.5 million bottles in 2012 to about 2.6 million recently, with strong seasonal cycles around October and December and May and June, and a drop in January and February.

The skills for these TVFs are available at the Google Skills GitHub repository, with documentation links for each function, including AI.KEY_DRIVERS, AI.CAUSAL_EFFECT, ML.CORRELATION, ML.SEASONALITY, ML.TREND, and ML.DETECT_CHANGE_POINTS, as well as BigQuery AI/ML support in Conversational Analytics.