Oracle adds an AI assistant for cloud management, with approval before changes

Oracle cloud customers can now ask an integrated AI assistant to investigate their infrastructure and propose changes, rather than navigate each resource separately. On October 26, 2026, Oracle made Console AI generally available to commercial Oracle Cloud Infrastructure customers. Supported changes require approval before the assistant applies them.

A conversation becomes a proposed configuration

Console AI lives inside the OCI Console, Oracle’s cloud management interface. Users describe a task, ask follow-up questions and review a proposed plan. Oracle says the assistant supports resource creation and management, configuration assessments, cost investigations and troubleshooting. It is not limited to answering documentation questions.

Oracle illustrates the workflow with a storage bucket for application release files. The user first asks for a private bucket and a proposed configuration, then follows up to enable versioning and assign an existing customer-managed encryption key from OCI Vault, Oracle’s key-management service.

The user reviews and approves changes along the way. A final request asks the assistant to read back the resulting settings: private access, versioning enabled and the assigned key. The example brings work across Object Storage and Vault into one conversation, including a check of the completed configuration.

Network fixes and cost reviews take different paths

For a security review, Oracle shows a user asking the assistant to inspect a virtual cloud network without making changes. Console AI identifies a rule accepting inbound traffic to every TCP port from any IPv4 address. It presents the finding, supporting details and recommendations together.

When asked to address that finding, the assistant proposes removing the unrestricted rule and explains which rules will remain. It applies the correction after approval, then displays the updated configuration. The initial inspection and the later request to change the network are separate steps.

A troubleshooting example runs in the other direction: a connection fails because an inbound rule is missing. The user asks for access on TCP port 8080 from an approved source range while preserving unrelated rules. Console AI proposes that narrow change for review, then checks the configuration after applying it.

Cost analysis instead points users toward further investigation. Oracle’s example ranks spending by service and suggests checking unattached storage volumes. Users can then ask for those volumes and review their ownership, age, tags and whether the data is still needed before deciding which are cleanup candidates.

Administrators can permit inspection without changes

Console AI operates within existing OCI Identity and Access Management permissions, the rules governing what a user may access. Oracle says it cannot view or manage resources outside those permissions. For resource-changing requests, the user can approve or cancel the proposed plan.

Tenancy administrators can add restrictions through deny policies. They can restrict Console AI access for a group or compartment, or allow inspection while preventing resource-changing actions through Console AI. These restrictions do not alter the user’s permissions through the regular console, command-line tools, software development kits or APIs.

Oracle still advises users to review responses and proposed changes before relying on them. Its launch examples illustrate intended workflows, rather than provide independent measurements of accuracy or time saved.

Customers can open Console AI from the console header or Help menu. It is currently available at no additional charge, although usage limits or throttling may apply. Oracle says pricing, availability and limits may change over time.