Industry change
Databricks outlines three ways to run OpenAI models on governed company data
- Published
- Source
- Databricks blog
Summary
Databricks published a blog post organizing three enterprise adoption patterns built on its multi-year partnership with OpenAI, formed in September 2025. They are: a Genie plugin in ChatGPT that answers questions using a company's approved metric definitions (Genie Ontology); routing Codex through Unity Gateway so leaders control budgets, logs, and model choice; and serving GPT models next to company data with Model Serving while deploying agents built with the OpenAI Agents SDK as Databricks Apps. The company says users only see data they already have permission to see. The original page shows no publication date, so the November 13 date follows multiple secondary reports.
Why it matters for our work
This lays out concrete architectures for running agents inside a company's real data and permission structure. It is a useful reference for letting non-analyst staff ask questions in the company's own definitions and for engineering groups controlling coding-agent costs and access.
Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.