Oracle announces Fusion Companion for business requests through ChatGPT

Oracle announced Fusion Companion on October 26 for requesting business work through interfaces including ChatGPT. It selects processes using objectives and permissions, letting employees delegate goals.

A new Fusion starting environment, Agentic Home, tracks delegated work, proposes corrective actions for exceptions and escalates matters requiring human judgment.

Oracle also announced supply-chain and customer-experience applications for reviewed plans and approved configuration changes. They are available within Fusion Cloud SCM and CX.

The supply-allocation application weighs competing demand against constraints involving finished goods, components, production lines and work centers. Planners can compare trade-offs, refine a plan and approve it. AI therefore enters the process of deciding how limited operational resources should be allocated.

Shipment-load planning considers timing, capacity, service requirements and cost when consolidating shipments. Procurement analysis examines supplier agreements and purchasing records for off-contract spending or contract leakage. It identifies causes and savings opportunities, with corrective actions subject to human review and approval.

Configuration changes pass through a review report

The quoting and pricing application checks proposed configuration changes and dependencies before migration from a test environment to production. It creates a reviewable report and identifies issues before deployment. Approved migrations execute within delegated authority.

A service-health application examines problems across knowledge, automation, routing and policies, proposing actions for review and approved execution. An executive-briefing application combines account relationships, product footprint and revenue opportunities into an editable document, updating it as account information changes.

Oracle also described financial applications for close readiness and reconciliation exceptions. Its ledger approach combines deterministic computation with authorization and approval controls. Workforce-orchestration plans likewise remain subject to organizational review and approval. A payroll workspace compares current pay, balances and costs with previous periods, identifies material variances and prepares recommended follow-up actions for approval.

The execution foundation, Fusion Claw, was introduced on September 29. At that point Oracle announced 25 Claw-powered applications within a portfolio of 75 agentic applications. The current additions include Companion, the Home interface and further applications for supply-chain and customer operations.

The existing runtime establishes objectives, operating procedures, policies, permissions, risk limits and escalation requirements. Automation levels depend on explicitly delegated authority. Oracle also described an audit receipt recording the evidence, decisions, actions, transactions and results behind each run.

Identity, policy, approvals and execution accountability remain in Fusion. The announcement specifies no ChatGPT subscription requirement, regional rollout date or separate price.

Organizations and partners configure applications through the existing Oracle AI Agent Studio. Its documentation describes no-code and pro-code creation, reusable agents and transaction-linked return-on-investment metrics. The business-value dashboard shows individual and portfolio impact using actual usage data, according to Oracle. This describes a dashboard capability, not a measured savings rate across the newly announced apps. Monitoring tracks agent performance and behavior. Listed capabilities include quality testing before deployment, approval workflows and oversight controls. The studio also provides direct access to Fusion transactions and data, along with interoperability for connections to external systems and other agents.