Oracle announced AI features across finance, HR, supply chain, sales, and service, including agentic applications and interfaces for delegating governed work.
Oracle announced new AI capabilities across Fusion Cloud Applications on October 26, positioning its software suite to let users delegate business objectives to agents and track progress. The release spans finance, human resources, supply chain, sales, and service, and includes 25 Fusion Agentic Applications powered by Oracle Fusion Claw.
Oracle describes the applications as tools for specialized business processes rather than general-purpose chatbots. Examples include a ledger application that investigates accounting exceptions and prepares authorized actions, workforce planning that weighs staffing options, and supply planning that compares ways to allocate scarce materials. Oracle says the applications can handle longer-running work and continuously replan, while using deterministic computation and operating within organizational controls. Several workflows prepare recommendations or actions for human review and approval.
The announcement also introduces Fusion Companion, an interface that Oracle says can interpret a user’s objective using business context, data, and permissions, then coordinate relevant transactions, workflows, approvals, agents, and applications. Fusion Agentic Home is described as a place for users to define goals, delegate work, and monitor its progress. Oracle says Companion can also use context from other enterprise systems and coordinate authorized actions across them.
The products are designed for organizations already using Oracle’s Fusion application environment, and the announcement does not establish that every feature is available to every customer or that the systems operate without supervision. Oracle’s framing emphasizes governance and accountability, but deployment teams will need to determine how permissions, approvals, and auditability map to their own policies and processes.
The broader move is toward AI that executes multi-step work inside systems of record, not just generates text alongside them. That could shift how enterprises evaluate application software, while making control over agent authority and the reliability of automated actions central procurement questions.