Oracle Ships 25 Claw-Powered Agentic Applications Across ERP, HR and Supply Chain, and Says an Early Hospital Customer Cut Configuration Work by 80 Percent

Oracle announced Monday at its AI World conference in Las Vegas that 25 agentic applications are now available across Oracle Fusion Cloud ERP, Human Capital Management, Supply Chain and Manufacturing, and Customer Experience, all powered by a new execution runtime called Fusion Claw. Alongside them, the company introduced Fusion Companion, a coordinating layer that can be driven through supported AI tools such as ChatGPT, and Fusion Agentic Home, a native AI workspace inside Fusion Applications. Chris Leone, Oracle's executive vice president of applications development, framed the shift as software that increasingly does the work while people define objectives and set policies.

Fusion Claw is the technical core of the announcement. Oracle describes it as a new runtime that supports larger scale and longer-running work, continuous replanning, deterministic execution and what it calls increased governed autonomy. The design intent is a "system of outcomes": people define the objective, the authority limits and the accountability, and the agentic application coordinates the process to reach the result, with approved actions executed inside the enterprise's own controls rather than in an external agent framework.

The 25 applications are aimed at specific finance and operations workloads rather than general chat. Ledger works across large volumes of accounting entries to reconcile information, investigate exceptions and apply deterministic computation to approved financial analysis. Workforce Orchestration develops staffing plans that balance demand shifts, cost and continuity, then coordinates approved actions. Payroll Activity Workspace compares pay and costs against prior periods to flag material variances. Supply Planning Allocation evaluates competing demands for scarce materials and capacity; Procurement Spend detects maverick spend and contract leakage and quantifies savings; others handle CPQ configuration migrations, service health checks and field demand planning. In each case the agent prepares recommended actions for human review before executing approved ones.

Fusion Companion extends the same machinery to wherever employees already work. It acts as an intelligent layer over Fusion Applications, using enterprise business context, data and user permissions to understand what a user wants to accomplish and to automatically select and coordinate the right transactions, workflows, approvals, AI agents and agentic applications. Users can invoke it through supported AI tools including ChatGPT, and it can coordinate authorized actions across other enterprise systems while respecting existing security and permission boundaries. Fusion Agentic Home gives the same capability a native home inside Fusion, letting users define objectives, delegate to agents and track progress without navigating applications.

The second announcement targets the less glamorous side of enterprise AI: implementations. Oracle AI Success Navigator, formerly Cloud Success Navigator, is now built on the Fusion Setup and Configuration Registry, a machine-readable source of setup definitions and product-defined best practices that Oracle maintains as reusable AI skills. Six specialized agents have been added: Discovery analyzes environments and identifies scope and risks; Design translates business processes into application designs; Configuration builds prototypes from those designs; Testing builds use cases and alignment steps; Integration scopes and generates integration code; and Innovation identifies adoption opportunities. Early adopter Lexington Health, a healthcare provider, says the agentic delivery approach cut certain design and configuration effort by 80 percent during its Fusion deployment.

What Oracle is selling is different from what the model labs are selling. Rather than stronger reasoning, the pitch is determinism: agents whose actions run inside the system of record, within policy limits, with every step attributable and reversible. That positioning matters for finance and HR workflows, where a plausible-but-wrong autonomous decision is worse than no decision, and it explains why Oracle insists on review-then-execute patterns rather than free-running agents.

The competitive read is that enterprise application vendors are converging on outcome-based agents as the next pricing tier, with Salesforce, SAP and Workday all pushing their own agent platforms. Oracle's bet is that regulated enterprises will pay for autonomy that lives inside the vendor's own transactional engine, with the vendor carrying the compliance burden. The open question, as always with enterprise agent announcements, is how much of the claimed autonomy survives contact with auditors, and whether the 80 percent configuration savings at a single early customer generalizes to the messy estates most corporations actually run.