BigQuery AI.AGG: summarisation, quality and cost
AI.AGG brings grouped text and image summarisation into BigQuery. Test the summaries as carefully as the query.
What happened
Google documents AI.AGG as a Gemini-powered aggregate function that returns a string for each input group, using natural-language instructions.
Why it matters
Teams can analyse unstructured material closer to existing SQL workflows. The result still needs evaluation against the underlying records.
Who it affects: Analytics engineers, support analysts and BigQuery teams working with unstructured data.
A practical example
For support-ticket themes, compare summaries against a manually reviewed sample and check whether rare but important issues disappear.
What to do next
Begin with a bounded dataset and a written definition of an acceptable summary.
- Record the instruction and input selection.
- Check source support for each substantive claim.
- Measure inference cost before increasing volume.
Signal Take
Adopt the function when it improves a specific analytical task, with quality and cost both visible.
Scope and limitations
Review current permissions, supported inputs, model availability and quotas in the linked documentation. No query was benchmarked for this article.
Sources and editorial record
Google Cloud Documentation (opens in a new tab)- Source type
- Product documentation
- Source checked
- 19 Sept 2026
Prepared with AI assistance and checked against the linked source. This is editorial interpretation, not an independent product benchmark. How we work.