Photo by Kindel Media on Pexels
For years, the telecommunications industry has been awash in data — petabytes of telemetry streaming from base stations, core networks, subscriber systems, and interconnects. The challenge was never really about collecting that data. It was about doing something meaningful with it. Now, Rakuten Mobile is making a compelling case that the next evolutionary step isn’t just smarter analytics — it’s agentic AI: systems that don’t merely observe network conditions but act on them autonomously, in real time.
The Shift from Insight to Outcome
The telecom AI conversation has long revolved around dashboards, anomaly detection, and predictive modeling. These tools deliver insight, but they still rely on human operators to translate that insight into action — a process that introduces latency, inconsistency, and scalability constraints. Rakuten Mobile is challenging this model with what industry observers are increasingly calling the “agentic network,” where AI doesn’t just flag a problem but resolves it.
At its core, an agentic network leverages AI agents — autonomous software entities that perceive their environment, reason about it, and execute decisions without waiting for human approval. In a telecom context, this means an AI agent might detect abnormal signaling patterns indicative of SIM-swap fraud, cross-reference subscriber behavior history, and trigger an account lock or network-level block — all within milliseconds, and all without a human in the loop.
This isn’t speculative. Rakuten Mobile, which operates Japan’s newest and most cloud-native mobile network, has been systematically building the data infrastructure and AI layer necessary to make agentic networking a practical reality rather than a PowerPoint concept.
Fraud Prevention as a Proving Ground
One of the most immediately tangible applications Rakuten has leaned into is AI-driven fraud prevention. Traditional fraud management systems in telecom are rule-based and reactive — they catch known fraud patterns but struggle with novel attack vectors. Rakuten’s approach integrates machine learning models trained on real-time and historical network data, enabling the system to identify behavioral anomalies that wouldn’t match any predefined rule set.
What makes the agentic framing significant here is the response layer. Rather than generating an alert for a security operations team to investigate hours later, the system is architected to initiate protective actions autonomously. This closed-loop design reduces the window of exposure dramatically — a critical advantage in an era where fraud techniques evolve faster than operations teams can update their playbooks.
RAN Energy Optimization: Where Automation Meets Sustainability
Perhaps the most technically intricate deployment of Rakuten’s agentic AI approach is in Radio Access Network (RAN) energy management. The RAN is the single largest consumer of energy in a mobile network, often accounting for 70–80% of total operational energy costs. For an operator running a nationwide network, even marginal efficiency gains translate to significant OPEX savings and carbon footprint reduction.
Rakuten’s cloud-native, Open RAN-based architecture provides a distinct advantage here. Because the RAN software stack is disaggregated and runs on standard hardware, it exposes APIs and data hooks that proprietary systems from legacy vendors typically do not. This openness allows AI agents to access granular, real-time performance metrics — traffic load per cell, interference levels, user distribution — and dynamically adjust power states, antenna configurations, and sleep mode schedules without human intervention.
The Open RAN Advantage
Legacy RAN deployments from vendors like Ericsson, Nokia, or Huawei operate largely as black boxes. Operators can tune certain parameters, but deep, real-time programmatic control is limited. Rakuten’s decision to build its network on Open RAN principles from day one — working through its subsidiary Rakuten Symphony to productize that architecture for other operators — means its AI layer has far greater surface area to work with. The RIC (RAN Intelligent Controller), a core component of Open RAN architecture, serves as the orchestration plane through which AI-driven xApps and rApps can issue control commands to the radio layer in near-real-time or non-real-time loops.
This architectural openness is not just a philosophical choice — it’s the technical prerequisite for agentic networking at the RAN level. Without disaggregation and open interfaces, AI remains a spectator rather than a participant.
Building the Data Foundation
Underlying all of this is a sophisticated data platform. Agentic AI is only as good as the data pipeline feeding it. Rakuten has invested heavily in unified data lakes that consolidate streams from the RAN, core network, OSS/BSS systems, and external threat intelligence feeds. This convergence allows AI models to reason across domains — understanding, for instance, how a congestion event in the RAN correlates with a spike in customer care calls or a drop in revenue-generating transactions.
The platform is designed for low-latency data ingestion and processing, which is non-negotiable when decisions need to happen in sub-second timeframes. Streaming analytics frameworks and event-driven architectures replace the batch-processing models that would make real-time agentic responses impossible.
Industry Implications and the Road Ahead
Rakuten Mobile’s agentic network vision arrives at a moment when the broader telecom industry is under intense pressure to reduce costs, improve service quality, and differentiate in commoditized markets. The operators that crack autonomous network management first will gain a structural cost advantage that compounds over time — requiring fewer NOC staff, responding faster to incidents, and optimizing resources continuously rather than periodically.
Through Rakuten Symphony, the company is actively commercializing its learnings, positioning itself not just as a Japanese MNO but as a global technology exporter. If the agentic network model proves out at scale, it could fundamentally reshape expectations for what intelligent network operations look like — and raise uncomfortable questions for operators still dependent on traditional vendor ecosystems that resist the openness agentic AI demands.
The data has always been there. Rakuten Mobile is making the case that the industry has finally built the tools to let it act.
