JevMade

Sign in
← Back to Jev News Releases & Updates

Databricks Introduces ai_decide for Governed SQL and Real-Time Data Decisions

Databricks launched native AI Function ai_decide on September 30, bringing sub-second structured classification, routing, and evals directly into SQL and governed REST endpoints.

1 min read By JevMade Source: Databricks Blog ↗
Editorial collage of a brushed-brass bank vault handwheel connected to steel storage cylinders controlled by an orange lever.

Databricks became the first major enterprise data platform to ship a native System One primitive on September 30, announcing ai_decide—a built-in Databricks AI Function engineered for fast, low-cost structured decisions directly over governed enterprise data.

Decisions in SQL and REST without moving sensitive data

Authored by Databricks engineers Jane Zhang, Matthew Ding, and Archika Dogra, the launch post positioned ai_decide for three high-frequency workloads: routing incoming prompts to the right downstream LLM, tagging customer reviews and documents with structured metadata, and running batch agent quality evaluations.

Because ai_decide runs natively inside the customer’s Databricks workspace—callable at batch scale in SQL or in real time over REST—enterprises avoid exporting regulated rows to an external API while sidestepping the latency and cost of general-purpose text generation.

Early signal for data-warehouse convergence

Databricks’ September 30 launch kicked off a rapid October wave across cloud data platforms, followed within weeks by Snowflake Decision and MotherDuck’s prompt_jev(). Together, they confirmed that structured decision models are becoming a standard SQL-era primitive alongside traditional joins, regexes, and aggregations.

Keep what you save

Sign in to keep bookmarks and private notes on every device. We email you a link; there is no password, and the same link creates your account.