OpenAI and Oracle agree to bring Oracle business apps into ChatGPT
Oracle AI Workspace is still in development. The agreement also includes security and app-building offerings, plus U.S. marketplace access to OpenAI services.
Loading page…
Oracle AI Workspace is still in development. The agreement also includes security and app-building offerings, plus U.S. marketplace access to OpenAI services.
Listen to this story
The expanded OpenAI–Oracle partnership spans both directions: Oracle is building a ChatGPT interface across its business applications, while adding OpenAI tools to security, small-business and app-building products. Oracle has not announced when AI Workspace will be generally available, so the cross-application workflows remain planned rather than a demonstrated customer result. U.S. customers can already buy OpenAI API, Codex and ChatGPT Work through Oracle Marketplace, using existing procurement processes and eligible Oracle credits.
Oracle says it has 130,000 active ChatGPT users and more than 95,000 active Codex users.
Oracle’s recruiting team reports preparing market research in 15–20 minutes with ChatGPT Work, versus two to four days previously; this is not a Workspace result.
Oracle AI Security Mesh will use OpenAI cyber models for vulnerability assessments, with Oracle running tests within customer-authorized scopes.
Oracle customers could use ChatGPT to investigate spending, pull together business information and complete approved tasks without navigating each Oracle application separately. OpenAI and Oracle announced an expanded partnership on October 27, 2026, pairing that planned integration with new Oracle offerings powered by OpenAI and a U.S. marketplace purchasing route.
Oracle is developing Oracle AI Workspace to connect its applications and data to ChatGPT. The intended experience starts with a person describing an outcome, rather than choosing a business application first. From there, the system would help find relevant information and carry out workflows approved by the organization.
OpenAI illustrates the idea with a new finance employee investigating costs and purchasing decisions before the monthly accounts are closed. That person could ask what remains to be done, drawing on information from finance, procurement and planning applications. It is an example of the intended workflow, not a reported customer result.
The companies say each organization’s permissions and approvals would still control access. They also aim to make administration and sign-in easier, so people and AI agents can work across business apps. General availability for Workspace was not announced.
The agreement also puts OpenAI models and tools inside Oracle offerings. Oracle announced three offerings alongside Workspace:
The companies’ commercial relationship includes the OpenAI API, ChatGPT Work and Codex. OpenAI said Oracle reported 130,000 active ChatGPT users and more than 95,000 active Codex users.
OpenAI pointed to Oracle's recruiting team, which it said now prepares market research in 15–20 minutes with ChatGPT Work, compared with two to four days before. This result comes from Oracle's own use; it is not an outcome measured for Workspace or the new product bundle.
Story updates
Oracle’s forthcoming AI Workspace would give AI responsibility for coordinating business tasks, while leaving Oracle applications to validate requirements, check approvals and make authorized changes. In the company’s design, an assistant could help resolve a delayed customer order across several systems without replacing the applications that control the underlying business transactions.
Oracle’s forthcoming AI Workspace would give AI responsibility for coordinating business tasks, while leaving Oracle applications to validate requirements, check approvals and make authorized changes. In the company’s design, an assistant could help resolve a delayed customer order across several systems without replacing the applications that control the underlying business transactions.
Oracle introduced Oracle AI Workspace on October 27, 2026, as a planned interface in ChatGPT connecting Oracle applications, enterprise data and authorized third-party tools. OpenAI confirmed that Oracle is developing the integration. The intended starting point is a business goal expressed by an employee, rather than a particular application the employee must open first.
Oracle illustrates the problem with a customer order threatened by a supplier delay. Addressing it could require finding an alternative product, checking inventory, understanding the customer’s requirements and validating transportation. The business would also need to assess the financial impact before updating the order. A single customer commitment therefore involves information and decisions spread across several functions.
In that scenario, Workspace would keep track of the original goal while coordinating the applications, AI agents, data and tools needed to address it. Oracle describes employees focusing on the outcome and the decisions requiring their attention, instead of manually connecting each step. This is an illustration of the intended product, not a reported customer deployment or measured result.
The proposed handoff is explicit: Oracle applications would validate business requirements and approvals, then perform the authoritative work—the changes that take effect in business systems. Existing rules, permissions and approval authority would remain in place. Information gathered across the enterprise could support a recommendation without automatically triggering action; where human approval is required, the responsible person would retain that decision.
OpenAI’s announcement supplies a different example: a new finance employee investigating costs and purchasing decisions before month-end close. The intended experience would bring together information from finance, procurement and planning applications, helping that employee identify what remains to be done. It describes the same goal-led interface through a financial investigation rather than an operational disruption.
Oracle also identifies potential uses where coordination, rather than a single answer, is the central task:
Supply-chain recovery: evaluate alternatives, assess operational and financial effects, and coordinate a recovery plan.
Employee onboarding: coordinate HR, IT, facilities and finance work covering access, equipment, training and other resources.
Workspace is “coming soon,” according to Oracle, which says it will share more about availability and capabilities closer to launch. The October 27 announcement establishes the proposed experience and division of responsibility, not current access to the product.
Oracle’s accompanying disclaimer also limits how customers should treat those descriptions. It characterizes the announcement as general product direction, says it is not a commitment to deliver functionality and warns against relying on it for purchasing decisions. Development, release, timing and pricing remain subject to change at Oracle’s discretion. The concrete next step is the availability and capability information Oracle has promised closer to launch.
Oracle is introducing a managed service designed to investigate security weaknesses, reproduce them and check whether repairs work. Announced October 27, Oracle AI Security Mesh combines OpenAI’s cyber models with Oracle’s testing tools. The company says customers control the testing scope and must approve remediation changes before the service applies them.
Oracle is introducing a managed service designed to investigate security weaknesses, reproduce them and check whether repairs work. Announced October 27, Oracle AI Security Mesh combines OpenAI’s cyber models with Oracle’s testing tools. The company says customers control the testing scope and must approve remediation changes before the service applies them.
The service targets applications and infrastructure running in customers’ own data centers, Oracle Cloud Infrastructure or other clouds. Oracle supplies the testing operation rather than asking customers to assemble the models, connect the tools and coordinate the investigations themselves. Its stated purpose is to turn potential weaknesses into evidence security teams can act on.
Oracle’s product walkthrough starts with an application and information the customer already has. The platform maps systems and their connections, then proposes a test plan for approval. Specialized AI agents investigate possible weaknesses and test how someone might exploit them. A separate agent repeats the test to check that the problem is real.
Oracle introduced App Composer at Oracle AI World on October 27, 2026, giving employees a managed cloud service to create, publish and share applications. The service is in limited availability. Oracle’s pitch divides responsibility: employees build tools for their work, Oracle runs the infrastructure, and IT controls access, approved data, policies and costs.
Oracle introduced App Composer at Oracle AI World on October 27, 2026, giving employees a managed cloud service to create, publish and share applications. The service is in limited availability. Oracle’s pitch divides responsibility: employees build tools for their work, Oracle runs the infrastructure, and IT controls access, approved data, policies and costs.
The announcement focuses on what happens after an employee makes something useful. Oracle gives the example of a finance worker building an invoice-exception tracker, then facing requests to share it with colleagues. That raises questions beyond creating the app: which business data it can use, who may open it, and who supports it as more people depend on it.
Employees start through an Oracle plugin inside an AI assistant. They can create an application there or bring in one they have already built, including a Java app.
Loading discussion...
The company illustrates the task with an ordering application that lacks an access check. A customer can sign in legitimately but retrieve another customer’s private order. In that example, the investigation tests whether the application exposes information to the wrong person. It is an illustration of the intended testing process, not a disclosed customer incident.
For findings, Oracle says the service supplies reproduction steps and test results. It uses context about the customer’s systems and business priorities to help rank the most critical risks. That gives teams both an explanation of how a weakness could be exploited and evidence they can use when deciding what to address first.
Coding agents can help prepare code fixes for the team’s approval. Within the customer-approved testing scope, the platform evaluates proposed fixes against the identified vulnerability and checks their effect on expected application behavior. Oracle also describes checking applied changes to confirm that the gap is closed and normal activity still works.
Oracle says agents’ discoveries stay in the business’s private knowledge base. Future tests can draw on those findings to connect clues and investigate deeper security gaps. The described workflow therefore carries information from one investigation into later testing, rather than treating each finding as an isolated result.
The operational safeguards address a different question from whether a vulnerability is real: what the investigating agents are allowed to do. Oracle says each agent has a distinct identity, a defined task scope and just-in-time permissions—access granted when needed. Independent monitors can detect and stop activity outside the approved boundaries.
For organizations interested in using it, Oracle’s current instruction is to contact the company to get started. The announcement does not specify pricing or a general-availability date. Its product-direction disclaimer also says functionality, release timing and pricing may change, so the described workflow should not be read as a contractual delivery commitment.
Publishing triggers the next step: Oracle provisions approved resources in the customer’s Oracle Cloud Infrastructure, or OCI, environment. The employee-facing creation tool and the cloud deployment are parts of the same service. Oracle says it manages deployment and ongoing operations, rather than leaving the creator responsible for running the application after sharing it.
IT determines who can use an app, which business systems it can access, and which policies and cost controls apply. Oracle takes responsibility for releases, monitoring and recovery—the operational work needed after an app has been published.
Oracle illustrates the intended work with three departmental examples. These are proposed uses, not reported customer results:
Finance: share an invoice-exception tracker using approved Fusion or NetSuite data.
Operations: build an application for managing requests and handoffs.
Marketing: create a launch workflow that helps colleagues keep deliverables on track.
Oracle also describes a path for applications that become important enough for an engineering team to take over. The app’s source, settings and history can move with it.
For access, Oracle directs interested customers to contact the company. Its announcement also carries a future-product disclaimer: the described direction is not a contractual delivery commitment, and feature development, release timing and pricing may change.
Join the conversation
Explain where a shared interface would help—or get in the way.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.