Databricks outlines three enterprise paths for OpenAI tools
Databricks published an enterprise adoption overview on November 13, 2026, describing three ways organizations can use OpenAI tools: business-user work, governed coding, and custom agents for business processes.
Key takeaways · 4
- 01
For analytics workflows, ChatGPT can query Databricks data and return visual results through Genie, with users limited to data they are permitted to access.
- 02
For coding-agent governance, Unity Gateway can route requests, set team budgets, and keep an audit log.
- 03
ChatGPT subscription support in Unity Gateway covers Go, Plus, Pro, and Enterprise plans, but requests remain subject to the caller’s subscription rate limits.
- 04
Custom agents can be built and evaluated in Agent Bricks or with the OpenAI Agents SDK, then deployed as Databricks Apps.
Business-user work with governed data
Databricks describes ChatGPT using Genie to interpret company metric definitions, find relevant data, query Databricks, and return visual results.[2] Genie Ontology stores approved metrics, business definitions, and relationships among company data, according to Databricks.[2] Users only see data they already have permission to access, and answers link to their Databricks sources.[2] OpenAI says the connection can also support analytics questions, SQL generation, dashboard or Genie-output summaries, and packaging insights for teams.[3] Databricks says ChatGPT Work can suggest next steps and people to involve, share findings through Slack or email, and take user-approved actions through connected tools.[2]
Coding through Unity Gateway
For coding, Databricks says Codex signs in with a Databricks credential while the OpenAI key stays in the Gateway.[2] Engineering leaders can set budgets, review spending by team, and choose which GPT model handles each task.[2] Databricks recommends beginning setup by installing its CLI and running `ug codex`.[2] Teams can also connect ChatGPT Go, Plus, Pro, or Enterprise subscriptions to Codex through Unity Gateway.[4] Databricks documentation says the Gateway does not store those subscription credentials; it forwards them with each request, which is subject to the caller’s subscription rate limits.[4]
Building custom business agents
The third path is to build and evaluate custom agents in Agent Bricks or use the OpenAI Agents SDK, then deploy agents as Databricks Apps.[2] Databricks says that when it serves the model, teams do not need to manage an external API key.[2] Its official agent app template includes a conversational REST API, an interactive chat interface, and MLflow evaluation code.[1] Unity Gateway routes model requests, sets budgets by team, and keeps an audit log.[2] Those controls give teams points to consider when deciding how to manage agent access, usage, and review.
Availability and what comes next
Availability for the ChatGPT-Databricks connection depends on the plugin, the user’s plan, and workspace settings; some connections require administrator setup or approval.[3] Databricks says Unity Gateway is generally available, while its documented SQL, sandbox, and web-search tools and Unity Gateway skills remain in beta.[1] Databricks says it plans separate posts on the three adoption patterns over the following weeks, with architecture, setup, and customer lessons.[2] For teams planning a rollout, checking plan and workspace requirements, and distinguishing generally available capabilities from beta tools, can help focus evaluation.
The overview separates business-user analytics, coding, and custom-agent work into distinct adoption paths. Professionals evaluating these options can use the stated access controls, credential handling, rate limits, and availability caveats to shape a pilot and identify what needs administrator or engineering review.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
14 November 2026
Databricks outlines three enterprise paths for OpenAI tools
