Agentic-AI Orchestration in O-RAN for Enterprise Networks: Autonomous Policy, Scheduling, and Spectrum Adaptation

Abstract

The rapid development of O-RAN has opened new possibilities for intelligent or self-optimizing enterprise systems. This paper proposes an Agentic-AI-based orchestration meant for fully autonomous policy management, scheduling, and spectrum adaptation within enterprise O-RAN contexts. The system proposes the insertion of persona-based AI agents into the RAN Intelligent Controller to interpret network intents into operational policies on the fly. Agents use large language model-based reasoning, predictive learning, and contextual data when coordinating power, resource, watch, and frequency assignments in real time. Adaptive scheduling is provided against traffic and interference changes through prediction analytics and LSTM-based forecasting. Spectrum adaptation is carried out against changes in environment and network conditions, so as to guarantee signal quality and reduce outages. The autonomous orchestration boosts efficiency in service delivery, reliability, and speed over manual intervention. The framework is purposely designed to be deployable at the scalable edge for enterprise networks which need low latency and high availability, such as smart manufacturing, health-care, and campus systems. Thus, Agentic-AI O-RAN orchestration furthers the vision of self-governing, intelligent enterprise networks that would be able to seamlessly adapt to dynamic communication demands.