Verizon and Ericsson expand AI automation across the RAN

Verizon is working with long-standing network supplier Ericsson to integrate artificial intelligence into radio access network (RAN) automation. The effort includes a nationwide deployment of Ericsson’s Intelligent Controller (EIC) by the end of the year and supports Verizon’s broader goal of increasing autonomous control over its network assets.

The deployment will allow Verizon to run AI-native rApps developed by both Ericsson and Verizon. These microservices-based applications operate in non-real time and provide operators with greater control over open RAN environments. A more advanced application type, known as xApps, operates in near real time.

Verizon highlighted a connected-mode inter-radio access technology handover method already implemented between its 4G LTE network and 5G standalone (SA) network for fixed-wireless access (FWA) devices. The company said the method removes the need for manual intervention when handovers fail.

Matt Butz, Verizon’s vice president of network planning, said the work has produced measurable results. Artificial intelligence has been used to control massive multiple-input, multiple-output (MIMO) antennas and steer traffic, resulting in an almost 9% improvement in downlink data radio bearer throughput and an improvement of up to 8% in uplink user equipment throughput.

The companies also plan to test and deploy additional rApps, including a cell anomaly detector, a root-cause explainer, and an anomaly general optimizer. Verizon said these applications will support its progression toward level-four network autonomy.

TM Forum defines level-four autonomy as decision-making based on intent-driven, predictive analysis, with the ability to perform closed-loop management of service-driven and customer-experience-driven networks through AI modeling and continuous learning.

Ericsson’s automation platform

EIC is part of Ericsson’s broader Intelligent Automation Platform (EIAP). The platform is designed to support multi-vendor and multi-technology environments across 4G and 5G RAN. EIAP is used by operators in several countries and also supports AT&T’s $14 billion network overhaul program.

The latest project builds on several recent initiatives between Verizon and Ericsson. These include a pre-standard “6G” field call in the 7 GHz band at a site in California, Ericsson’s role in providing international private 5G services for Verizon Business, and the company’s participation in Verizon’s 6G Innovation Forum, which launched last year.

Verizon develops AI agents for network operations

Verizon CTO Yago Tenorio previously described the company’s AI RAN work while discussing its network automation plans. He said Verizon was developing specialized AI agents to handle increasingly complex automation tasks, including identifying the root cause of network underperformance, explaining the issue, and proposing a solution.

Tenorio said the company was using real-world tasks to refine and train the AI systems. The next stage would involve moving the agents from general development into specialized roles using different tools within a prototype system, with the aim of putting that system into production within three months of his comments.

He also said the classification of the system by autonomy level was less important than its structure. The planned architecture would consist of agents and sub-agents monitoring the network continuously, identifying patterns and anomalies, and assigning specialized sub-agents to conduct root-cause analysis and develop solutions autonomously.

According to Tenorio, early testing had reduced some tasks that previously required hours to approximately 90 seconds. He said Verizon wanted the work in production before the end of the year, initially focusing on the radio network before expanding to transport and core network functions. He expected the agents to begin supporting operations within six months.

Verizon has also worked with Samsung and Qualcomm on a multi-vendor AI-RAN deployment using management applications running on a RAN intelligent controller (RIC). That project included a Samsung-developed rApp designed to manage network efficiency.