How to put generative AI directly into a dbt pipeline with plain SQL

Google Cloud’s Alicia Williams demonstrated at dbt Summit 2026 how BigQuery AI functions can embed generative AI directly in SQL and dbt pipelines, avoiding separate Python services or API workflows. BigQuery’s AI.GENERATE can process table rows with LLM prompts, while structured output schemas provide typed JSON for dependable downstream use; managed functions including AI.IF, AI.CLASSIFY, AI.SCORE, and AI.AGG support common filtering, categorization, scoring, and summarization tasks at lower cost. Optimized modes for some functions use a small LLM-labeled sample to train lightweight distilled models, improving speed and reducing expense at scale. For production use, Williams recommended incremental dbt models to ensure only new records invoke AI, along with token estimates and caps, error thresholds, output schemas, and dbt tests to control costs and catch unreliable results. Her demonstration used a pet e-commerce dataset to join reviews with orders, score sentiment across several categories, and produce product-level summaries that could help managers identify issues such as faulty product components.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.