Oracle adds Fusion Claw to automate complex business tasks

Oracle announced today that its Fusion suite now includes a feature called Fusion Claw. The runtime lets the software act as an AI‑driven agent that can plan, model and execute longer‑running business processes. It combines large language models – the kind of AI behind ChatGPT – for reasoning with Oracle’s traditional transaction engine for calculations. The company says the first 25 Claw‑powered applications are available immediately and are included at no extra charge with the existing agentic applications product.

Claw is aimed at tasks that need repeated adjustments, such as creating staffing schedules that respect employee qualifications, labor rules and cost targets. The agent can propose options, run the necessary math, and, if given permission, update the official records. An “Outcome Receipt” records the policies applied and the decisions made, providing more detail than a normal audit log.

How does Fusion Claw differ from earlier agents?

Earlier Oracle agents focused on narrow goals like boosting subscription renewals. Claw is built for multi‑step work that may need to be re‑run when conditions change – for example, reopening a schedule if someone calls in sick. The architecture limits the use of expensive AI models: the language model is only invoked for reasoning, while bulk data processing stays in Oracle’s conventional engine. The initial release works with Google’s Gemini and OpenAI’s models, though customers cannot yet plug in their own lower‑tier models.

What this means for customers

Businesses can set an “Enterprise Operating Envelope” to define objectives, risk limits and approval steps. They can choose a mode where a human reviews each plan or let the agent act autonomously within the defined boundaries. Oracle expects companies to start with human oversight and expand automation as they gain confidence. The company did not share specific cost‑saving figures or performance data from real‑world deployments.