Multi-agent collaboration and digital twins are the route to O&M transformation through Level 4 autonomy

Commercial context

Telecom networks have undergone a fundamental shift over the past decade. The industry has moved from a network-centric approach to an experience-led approach, driven by user expectations rising from simple availability to an expectation of zero defects. In pursuing this aim, CSPs struggle with three core limitations: limited end-to-end user experience assurance, the absence of proactive fault prevention, and insufficient automation across domains.

Operational challenges reveal how deep these issues run. Delayed user perception assurance remains widespread, with teams responding reactively to symptoms instead of predicting issues in advance. Manual, lengthy complaint-handling processes create bottlenecks that slow resolution and damage customer satisfaction. Fault analysis is also often imprecise with work orders that lack clear metrics on how faults affect user perception. All faults appear equal, and redundant tickets stack up. The phenomenon of ‘one fault generating multiple orders’ also exists, further adding to the challenge.

Meanwhile, fault handling is inconsistent and inefficient. Root cause analysis can be slow, troubleshooting still depends heavily on expert intuition, and systems across domains do not collaborate automatically. Without simulation or validation tools, solution implementation risks also rise. Together, these issues extend mean time to repair (MTTR) and cause unpredictable operational risks. They also result in poorer network quality and reduced customer retention while creating high O&M costs.

Successfully addressing these issues presents an obvious commercial opportunity. Reducing manual operations can significantly cut operational expenditure, while pre-emptive and accurate cross-domain fault resolution dramatically improves network reliability. By strengthening end-user experience and reducing downtime, operators can turn dissatisfaction into trust and loyalty.

The solution

To this end, the ‘Agentic AI for customer-centric O&M’ Catalyst proposes a new approach to tackling these challenges through the integration of signalling analysis large models, spatiotemporal analysis large models, multi-agent collaboration, and digital twin technology.

At the core of the solution is a multi-agent collaboration (agentic AI) technical architecture designed to support intelligent, automated fault identification and management. The system is built on a layered and modular framework composed of a perception layer, intelligence layer, application layer, and resource layer. Across these layers, multiple specialized agents communicate and collaborate through the communication-specific A2A-T agent protocol. This allows data to flow freely across domains that were previously siloed.

The Catalyst advances several innovations that set it apart. Its proactive, customer-centric precision capability uses signalling-analysis and spatiotemporal-analysis large models to pinpoint network quality issues and complaints before users notice them. The solution thus ends reactive firefighting while operating in alignment with IG1190 AIOps standards. By shifting the operational paradigm from ‘network-centric’ to ‘customer and business-centric’, the solution aims not only to resolve faults but to prevent them, thus preserving service quality and user experience. Seamless cross-domain automation is achieved through multi-agent systems using TMF640 fault management APIs, enabling end-to-end fault and complaint resolution that eliminates redundant tickets and delivers minute-level response times.

Meanwhile, risk-controlled optimization is made possible through digital twins: AI-generated solutions undergo simulation and stress testing using the network-simulation interface, using digital twin technology to verify resilience and performance before real-world deployment. This ensures they meet required standards, including compliance with TMF/DTOps guidelines, significantly reducing MTTR and operational errors. Additional TM Forum assets used include IG1218F Autonomous Networks Framework and TMF638 Service Inventory Management, IG1501 AN L4 Fault Management Solution Package, and TMF921 Intent Management.

Application and wider value

The Catalyst showcases how intelligent automation can transform O&M. Its benefits extend beyond cost reduction or faster resolution. It provides a foundation for CSPs to deliver consistent, customer-centric reliability while preparing for future network complexity.

The Catalyst delivers a broad range of measurable benefits. It drives major O&M efficiency upgrades by shifting operators from passive response to proactive prevention. Complaint-handling time drops from 15-20 hours to just four. Meanwhile, fault localization accuracy reaches 90% (20-30% higher than traditional methods), and automation rises from about 30% to 73%. Redundant work orders fall by 20%, cross-domain fault closure reaches minute-level execution, and MTTR drops by 15%. It also delivers replicable cost optimization with 32% investment savings. Furthermore, 3,700,000 yuan cen be saved in monthly labor reduction, and 4,440 person-days saved each month.

Overall, the Catalyst provides a transformative, customer-centric model for automated, predictive O&M, giving CSPs a practical path to Level 4 autonomy and a scalable blueprint for boosting satisfaction, efficiency, and competitive strength.