Microsoft and Oracle have announced integrations designed to let AI agents work with Oracle business data and Microsoft workplace context. Announced at Microsoft Ignite on November 17, 2026, the capabilities bring Oracle AI Database information into workflows involving Microsoft Excel, Teams, Copilot and Microsoft Foundry. The companies say the integrations are intended to help users and developers analyze and act on enterprise information without routinely copying it into separate AI systems.
The announcement spans business-user access, IT operations and agent development. It builds on Oracle AI Database@Azure, the jointly offered service that places Oracle database workloads on Azure infrastructure. Microsoft says the service is now available in 33 Azure regions. The companies frame proximity between databases and applications, along with connections to Microsoft AI services, as a way to support performance-sensitive workloads and make enterprise data available to AI-driven workflows.
Oracle data in familiar work tools
An Oracle AI Database plug-in for Excel is available now, according to Microsoft. It lets users connect Excel directly to an Oracle Database and bring trusted data into a worksheet, where they can use Copilot to reason over it. The stated goal is to reduce the need for manual export and preparation or for users to request that technical staff extract and format information.
Microsoft describes workflows in which Copilot analyzes Oracle transaction data alongside context from Work IQ. Work IQ draws on information in meetings, email, documents and Teams. In the examples provided, that combination could help analysts and business teams identify exceptions, explain changes in margins or demand, investigate reconciliation issues and prepare a plan related to supply-chain disruptions. Proposed actions remain subject to review, the company says; the announcement does not describe these examples as autonomous production actions.
The approach links structured records held in Oracle databases with work-related context held in Microsoft services. That distinction matters for enterprise use: database transactions can provide a record of business activity, while communications and documents may help explain the circumstances around it. The announced integrations aim to make both available in the tools employees already use, rather than requiring a separate dashboard or chatbot for each task.
Operational visibility and access controls
For IT teams, Microsoft announced a limited preview of Azure Workload Center for Oracle. It organizes a workload’s architecture, resources and dependencies around a construct called a Workload Virtual Instance, or WVI. Microsoft says the WVI is intended to simplify management compared with handling each component independently.
The preview integrates the WVI with Azure Monitor Health Models and the Microsoft Azure Copilot Run Oracle agent. These components are intended to connect operational signals across application, database and infrastructure layers. In a scenario described by Microsoft, an administrator investigating slow order processing could use a unified workload-health view and Copilot’s summary to identify areas for investigation. The administrator could then prompt multiple agents to coordinate an investigation and combine their findings.
The companies also emphasize controls over what agents can access. Oracle AI Database@Azure supports Microsoft Entra authentication and Oracle Deep Data Security, while database calls from agents run through a managed Model Context Protocol (MCP) server. Microsoft says users’ existing permissions constrain access, and actions are captured for audit. These mechanisms are intended to let organizations use agent-assisted investigation without granting agents unrestricted database access.
Developing agents with Oracle data
Developers can use Microsoft Foundry to build agents connected to Oracle AI Database through managed MCP servers and the agent-to-agent (A2A) protocol. The MCP servers provide governed access to databases running in Oracle Cloud Infrastructure or through Oracle AI Database@Azure. According to Microsoft, agents can use approved SQL and PL/SQL operations, tools and predefined reports; application roles determine which tools a user or group is permitted to use.
Microsoft presents Oracle AI Database as a potential source of knowledge, memory and data for Foundry agents. One example combines an agent that retrieves account and order information with product and finance agents that assess cross-selling opportunities and account health. The coordinated agents could produce a sales brief delivered through an API. Another example uses Foundry IQ and Oracle-backed memory to help a customer-service agent answer questions using policy information, account details and relevant customer history.
These are examples of intended applications, rather than reported customer outcomes. The announcement does not provide performance benchmarks for the new agent integrations or identify specific customers using them. It does note separate Oracle AI Database@Azure capabilities announced at Oracle AI World 2026: FedRAMP Class D certification was described as coming soon for U.S. government and public-sector workloads, and Proximity Placement Groups were said to deliver approximately 50% lower VM-to-Exadata network round-trip latency than deployments without PPG alignment.
Together, the announcements position the Microsoft-Oracle relationship as a way to connect cloud-hosted database systems with productivity tools, operational monitoring and agent-development services. The practical value will depend on how organizations configure permissions, validate agent outputs and integrate the capabilities with their existing database and application operations.