Work with Oracle Fusion Claw
Oracle Fusion Claw is a governed agentic execution runtime for agentic applications in AI Agent Studio. It combines AI reasoning with deterministic enterprise computation to complete highly complex work cost-effectively at scale.
Fusion Claw enables agentic applications to do more than provide recommendations. They can reason, plan, continuously adapt, and carry out complex business work using AI-driven reasoning, precise calculations, and authorized actions.
- Intelligent planning: Uses AI reasoning to analyze business objectives, context, and constraints, evaluate available options, and create a plan to achieve the desired outcome.
- Deterministic execution: Uses a separate execution layer to carry out authorized actions consistently and predictably at scale, reducing the need for repeated AI reasoning and optimizing execution costs.
- Continuous learning: Adapts to changing conditions during execution and uses insights from previous outcomes to improve the efficiency and reliability of recurring work.
Fusion Claw supports different operating modes, ranging from AI-assisted execution to fully automated execution within explicitly delegated authority. Users retain control by defining objectives, decision rights, policies, constraints, risk thresholds, and approval requirements that govern how Fusion Claw operates.
With Fusion Claw, you can describe a desired business outcome in natural language. The agentic application analyzes your objective, evaluates available options, and creates a plan using the capabilities of your Fusion applications. It can then reconcile data, perform calculations and simulations at scale, optimize the plan based on business constraints, and carry out the actions and transactions it's authorized to perform.
Fusion Claw operates within your organization’s established policies, permissions, approvals, risk thresholds, and authority limits. It can take actions within its delegated authority, request approval when required, and escalate exceptions when necessary. AI can plan and recommend actions, but it can’t grant itself authority to act. Each outcome run provides an auditable, traceable outcome receipt that records the decisions, actions, and results.
Application designers configure autonomous outcomes to define the goals, skills, capabilities, and agents that determine what Fusion Claw can do. For more information, see Configure Autonomous Outcomes.
After an autonomous outcome is configured, you can define an outcome and use Fusion Claw to achieve the desired business result. For more information, see Define an Outcome.