Proactive, predictive, and autonomous: genAI anomaly detection for RAN

Commercial context

Radio Access Networks (RAN) are becoming more complex, distributed, and data-heavy — but fault management processes haven’t kept pace. When anomalies arise, network teams must often rely on reactive alerts, manual root cause analysis (RCA), and time-consuming interventions. This slows down response, reduces reliability, and risks customer experience.

As networks scale, these inefficiencies become harder to manage. The cost of downtime rises, as does the pressure to meet performance targets in real time. For CSPs, managing this complexity without increasing operating costs requires smarter, more autonomous tools.

The GENAR – GenAI-enabled anomaly detection for RAN Catalyst addresses this challenge. It combines generative AI, cloud-native architecture, and autonomous intent-based workflows to detect, explain, and resolve anomalies before they disrupt service.

The solution

The GENAR Catalyst integrates Capgemini’s EIRA framework with Amazon Web Services technologies to provide a tightly coupled, GenAI-powered fault management system. The result is a scalable platform that spans the full anomaly management lifecycle — from detection to prevention. The solution uses EIRA’s AI-driven event correlation engine, enhanced with Amazon SageMaker’s MLOps for spatial-temporal analysis. This enables real-time anomaly detection, with models that continuously learn from data patterns across the RAN environment. Root causes are identified automatically, reducing noise and narrowing in on actionable faults.

Amazon Bedrock brings generative AI into the process. It simulates potential future anomalies, explains current incidents, and offers contextual insights — transforming raw detection into meaningful decision support. These insights power Aira’s GenAI-based App Gen tool, which builds and deploys corrective apps based on intent. These apps address root causes, reduce recurring faults, and prevent future disruption.

Built entirely on AWS services, the solution includes S3 for scalable storage and Glue for data integration. This allows rapid deployment, consistent updates, and future-ready scalability. The platform supports CSPs on their journey to Level 4 autonomy. It not only detects faults, but responds to them with minimal human input — enabling true self-adaptive, closed-loop management.

GENAR aligns with key TM Forum assets, including IG1343 (GenAI for fault detection), IG1373 (self-healing use cases), and GB1059 (Autonomous Network Level Evaluation Tool). These frameworks enable consistent benchmarking and help CSPs accelerate their path to autonomy.

Wider application and value

The operational gains are immediate. CSPs can detect faults faster, reduce manual RCA tasks, and automate resolution. This cuts mean time to detect (MTTD), improves uptime, and protects service quality during periods of high demand. Proactive detection means fewer outages. Networks resolve problems before users notice them. Customer satisfaction improves, churn reduces, and operational KPIs rise. Telefónica, a key partner in the Catalyst, has highlighted the project's value in boosting efficiency while maintaining high service standards. As explained by their Manager of Autonomous Network & Operations Transformation, Pedro Garcia Parra: "advanced anomaly detection ensures proactive management, enhancing customer experience, cost efficiency, and competitive advantage for CSPs." The wider industry could feel the key advantage of cost control. By reducing the need for manual investigation and cutting response times, the Catalyst helps CSPs lower operating costs without sacrificing quality. Teams can reallocate resources to strategic tasks, rather than firefighting daily incidents.

The solution supports a broader shift: from reactive monitoring to real-time observability. By covering the full fault lifecycle — from detection to prediction to adaptation — CSPs gain deeper visibility and faster control. The system scales across distributed environments, making it ideal for next-generation RANs and evolving 5G deployments.

For the public, and everyday life, the result is more stable connectivity, fewer dropped calls, and fewer service disruptions. As reliance on mobile networks increases — especially in areas where fixed infrastructure is lacking — network stability becomes a social need as well as a commercial one. By enabling intent-driven, generative AI-supported anomaly detection, the GENAR Catalyst demonstrates how CSPs can build smarter, more resilient networks. Now, there's an established and practical route to autonomous fault management — a critical step toward consistent, high-performance mobile service.