Oracle announces approved workflows for ChatGPT

At Oracle AI World in Las Vegas on October 27, 2026, Oracle announced Oracle AI Workspace as part of an expanded partnership with OpenAI. OpenAI president Greg Brockman described the partnership’s broader aim in the same announcement as “putting frontier AI into everyday operations.” The planned product would connect ChatGPT with Oracle applications, enterprise data and cloud services so employees could pursue business goals across systems.

In OpenAI’s partnership account, Oracle AI Workspace is a product under development: employees would describe what they need to accomplish, find relevant information and complete approved workflows in ChatGPT. Organizational permissions and approvals would still govern access and actions. The companies have described an intended experience, rather than a generally available service that employees can use today.

From a ChatGPT request to an Oracle action

Oracle’s product account illustrates the workflow with a customer order threatened by a supplier delay. Addressing that problem could require checking the customer’s requirements, finding an alternative product, confirming inventory and transportation, and assessing the financial effect before changing the order. It is an example of the intended experience, not a reported customer deployment.

The proposed path has distinct stages. An employee starts with the desired outcome in ChatGPT; AI Workspace brings together relevant information from connected applications and data sources; Oracle AI agents and applications help coordinate the task; and the underlying Oracle applications validate requirements, apply approvals and make authoritative changes. Keeping those stages distinct matters when a request moves from gathering facts to changing a business record.

In the supplier-delay example, locating available inventory could inform a recommendation without changing the customer order. A proposed substitute might then need to satisfy business rules and, where required, receive a decision from the person with approval authority. ChatGPT would help carry the goal across the task, while the application responsible for the order would make the final update under the organization’s controls.

That arrangement could spare an employee from reconstructing the same problem in several applications. It also makes the boundary of the AI role clearer: bringing information together and coordinating a proposed response are different from granting permission to act. The published description does not specify every handoff an employee would see, so the sequence explains the intended division of responsibilities rather than a finished interface.

Which systems the workspace would reach

The named connections are Oracle Fusion Applications, NetSuite, Oracle AI Database and Oracle Cloud Infrastructure, alongside other enterprise tools and data sources. AI Workspace is also intended to bring authorized third-party tools into the ChatGPT experience. The word “authorized” limits the scope: describing a goal would not itself give an employee access to a tool or information outside their organization’s permissions.

Those connections serve different parts of the proposed workflow. Business applications hold processes and records; data sources provide context; agents and connected tools can help coordinate work across them. An order problem, for instance, may depend on customer requirements, inventory, logistics and financial information before an authorized change is possible. Bringing that context into one task is the product’s proposed advantage, while each connected system retains a role in governing what happens to its records.

A separate example concerns a finance employee investigating costs and purchasing decisions before month-end close. The intended experience would draw information from finance, procurement and planning applications into the same task. That is a narrower claim than saying every Oracle customer already has those applications connected or that every third-party tool will be supported at launch.

Where approval remains—and what is available now

Existing business processes, rules, permissions and approval authority are meant to remain in place. Information from several systems may support a recommendation without automatically triggering an action. When a workflow requires human approval, the decision stays with the responsible person; Oracle applications apply the business rules and perform the authoritative change. A ChatGPT response alone would therefore not be the final record of an approved business transaction.

AI Workspace was introduced alongside other offerings in the expanded partnership, but their functions have different purposes. Oracle App Composer concerns building and deploying applications with Codex and approved Oracle data. Oracle AI Security Mesh concerns authorized security testing. AI Workspace is the proposed entry point for employees pursuing approved work through ChatGPT; the other offerings’ capabilities should not be treated as features of that workspace.

AI Workspace is described as coming soon, with its launch date, final capabilities and standalone price still unspecified. The next meaningful milestone for employees and administrators is a release that identifies supported connections and shows how permissions and approval paths operate in the available product. Until then, the practical change is a defined route for a business request to reach Oracle systems through ChatGPT, with consequential actions still dependent on the controls that govern those systems.