BT: Streamlining Network Operations with Digital Twins & Agentic AI

BT, AWS, and Celfocus have joined forces to deliver a unified assurance platform that correlates performance, fault, and topology data in near real time.

This solution predicts anomalies, accelerates root-cause detection, enables proactive network healing, and reveals customer impact during network faults. This initiative is transforming BT’s 4G/5G CORE and RAN assurance through graph-based analytics and AI-driven automation.

The Challenge

BT faced disconnected network tools and siloed KPIs that caused alarm fatigue, reactive troubleshooting, and slow fault isolation, with limited visibility across OSS domains and vendors.

Manual root-cause analysis prolonged Mean-Time-To-Repair (MTTR), while the lack of correlation between performance degradation and service impact hindered prioritisation. Repetitive, low-value alarms and reliance on costly, domain-specific tools further reduced operational efficiency.

Facing the same challenge?

Let’s Talk

The Solution

BT, Celfocus, and AWS delivered an integrated solution combining advanced analytics and AI-driven automation.

Multivariate KPI models built with Amazon SageMaker and AWS Glue enable early anomaly detection. Agentic AI-driven reasoning over an AWS Neptune graph supports rapid, cross-domain root-cause analysis by identifying causal nodes and dependencies.

The platform delivers end-to-end Anomaly Detection, Root Cause Analysis, and Service and Customer Impact Analysis, linking network events to services and customers via explainable AI, enabling BT's faster and accurate incident communication.

Benefits

The solution offers:

100% L1/L2 reduced MTTR: AI-powered anomaly detection automates 100% of temporal anomaly scoring across datasets, accelerating issue identification and reducing MTTR.
-50% proactive service management: predicts performance degradation before SLA breaches occur, enabling proactive intervention and protecting service levels.
-30% dark NOC operational efficiency: streamlined workflows and reduced dependency on domain tools lower OPEX and minimise the need for human intervention.
Vertical & horizontal causation: AWS Neptune and Agentic AI deliver precise RCA by identifying vertical causation within the same datacentre and horizontal causation across datacentres.

Project in a nutshell