Oracle launches governed AI agents for tasks in its Fusion applications
Oracle has introduced Fusion Claw, a runtime for governing AI agents that carry out work in its Fusion business applications on Oracle Cloud Infrastructure. The system aims to move AI from assistance to execution in processes such as finance and supply chain management, while applying controls to what agents can do.
Analysts say the approach could help tackle complex operational tasks and avoid the cost of sending huge volumes of business data through large language models. Oracle says agents can set up the work once, then hand calculations to deterministic computing engines within isolated containers and policy limits. The product can use models from Google and OpenAI, but adoption is still at an early piloting stage, and customers may become more dependent on Oracle’s applications and cloud.
- Oracle launched Fusion Claw to govern AI agents in its business applications.
- It separates AI reasoning from large-scale calculations to manage costs.
- Analysts cite early adoption and Oracle cloud lock-in as concerns.
New here? Start with this
Oracle is a large technology company that provides software and computing services to organisations worldwide. Its main products help businesses manage everyday tasks like accounting, inventory and customer relationships.
The company has now created a system called Fusion Claw that allows artificial intelligence to carry out work within these applications. Rather than simply suggesting actions to workers, these AI agents can now execute complex business decisions such as approving expenses or planning supply chains, whilst operating within rules set by the organisation.
This approach could help organisations complete work more quickly and cheaply, since complex decisions could be made without sending all company data through large artificial intelligence systems, which is traditionally costly and raises privacy concerns. The technology remains at an early testing stage with a limited number of customers.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Oracle's approach to governed AI agents represents a thoughtful step forward in business automation. By containing agents within policy limits and isolated containers whilst supporting multiple LLM providers, the system attempts to balance efficiency gains with necessary safeguards. The potential to reduce costs through deterministic computing rather than routing all data through expensive LLMs, combined with genuine operational improvements in finance and supply chain tasks, suggests this governance model could benefit enterprises that implement it carefully. The multi-provider approach also mitigates some vendor concentration risk.
The case against
Whilst the governance framework sounds promising, this remains an early-stage technology with significant risks that warrant caution. Enterprises deploying autonomous agents in critical business processes may find themselves increasingly dependent on Oracle's proprietary ecosystem, creating genuine lock-in risks. The complexity of truly governing AI agent behaviour in unpredictable real-world scenarios is not yet proven at scale, and the incentive structure (Oracle benefits from deeper customer dependency) raises questions about whether governance claims will hold up under commercial pressure. Customers should be wary of being early adopters of technology that could reshape their operational dependencies.
Read the full article at the source →
Originally published by The Register as “Oracle lets AI agents do the work, provided you stay in Big Red’s world”.