Understanding the Big Picture and Analyzing Trends with BigQuery's AI.AGG Function

BigQuery has announced the AI.AGG() function, which is now in preview. With this function, you can summarize and synthesize unstructured data or multimodal data spanning millions of rows using natural language instructions within a single line of SQL query.

While existing AI features have focused on analyzing individual data rows, AI.AGG() specializes in finding patterns in large-scale unstructured data such as logs or documents. For example, the following questions become possible:

• What are the top 3 feature requests among negative product reviews?
• What types of errors do users encounter most frequently, and how should they be investigated?
• What are the specific scenarios where automated agents fail to resolve customer issues?

It particularly demonstrates powerful performance in system log analysis. Instead of manually investigating thousands of log messages, warnings, errors, and stack traces one by one, you can use AI.AGG() to group and prioritize logs for efficient analysis. This makes it possible to identify even potential inefficiencies that occur without causing critical errors, such as memory thrashing or clock drift.