Login
Register
Summary (teaser): This Catalyst shows how a multi-agent AI system, intent-driven automation and TM Forum standards can help CSPs move from reactive, ticket-driven network operations to reliable, service-aware autonomy across the radio, transport and core.
Agentic NOC: AI-native operations for the autonomous telco
As communications service providers move toward autonomous operations, one of the biggest barriers is not a lack of automation, but the challenge of making automation reliable, explainable and useful in day-to-day operations at scale. The Agentic NOC: AI-native operations for the autonomous telco Catalyst project explores how an AI-native network operations center can help CSPs shift from reactive, manual processes to predictive, intent-driven operations that continuously learn from network, service and customer context.
The project was named winner of the Autonomy Accelerator category in the Open Innovation Catalyst Awards 2026, recognizing its contribution to accelerating practical progress toward autonomous networks.
Today's network operations center often works like a war room. A single incident can generate thousands of alarms across the radio, transport and core, which engineers still correlate by hand before forming a bridge call to fix it, a process that can take hours. Even modern AIOps tools tend to stop at a recommendation, leaving people to drive every step. The result is slow, costly and reactive operations at exactly the moment CSPs are adding more 5G services, edge computing and AI-enabled capabilities.
Agentic NOC addresses this by bringing together agentic AI, intent-driven automation and a shared data foundation. Specialized agents for fault correlation, anomaly detection, and service, business and customer impact work together as an agent of agents over a shared Snowflake data foundation, with data brought in and mediated by DigitalRoute. The aim is a single operating model where autonomous agents can detect, reason and act within clear guardrails and human oversight, rather than simply raising another ticket.
The Catalyst demonstrates how an AI-native NOC can restore service across the radio, transport and core in under 60 minutes, with no ticket, no bridge call and no customer impact in the typical case. It does this through two connected use cases running on real operator data. The first is intelligent fault correlation across radio and transport, led by champion T-Mobile USA. The second is transport and IP anomaly detection for mobile and fixed-wireless networks, led by champions Turknet and Axian.
In the first use case, radio fault and OSS data is brought into the shared data foundation, where the agents combine machine-learning correlation with topology reasoning to work out whether an apparent radio fault is in fact caused by an underlying transport problem. In the second, anomaly-detection agents built by Enfec learn normal performance-counter behavior and flag deviations early, before they turn into hard faults. Both cases feed a single operator dashboard that links technical events to business outcomes such as revenue at risk, penalty exposure and SLA breaches, in one view.
According to the project team, the Catalyst marks a shift from ticket-driven firefighting to intent-driven, closed-loop autonomy. Business and technical intents, such as SLAs, policies, and cost and energy targets, become the control inputs. The agents then sense, reason and act to keep the network aligned to those intents, rather than reacting to individual alarms one at a time.
Because the agents reason jointly across radio, transport and core, a symptom in one domain is traced back to its true cause in another, which removes much of the cross-team escalation that slows resolution today. The human role moves up from find and fix to setting intent and governing. People define the policies, risk appetite and guardrails the agents work within, and focus on the harder, ambiguous cases and new services.
The Catalyst is built on and aligned with a broad set of TM Forum assets. These include TMF921 Intent Management and the TMF921 Intent Management API, TMF915 AI Management API, TMF642 Alarm Management API, TMF656 Service Problem Management API, TMF701 Process Flow Management API and TMF702 Resource Activation API.
It also draws on TM Forum's Open Digital Architecture and Open APIs to design integration across components, together with the Autonomous Networks Architecture, the Autonomous AI Control Loop and the IG1220 closed-loop model. Supporting guides include IG1218 Autonomous Networks Technical Architecture, IG1339 Focus on Value: Operators' Key Requirements for Autonomous Networks and IG1501 AN Fault Management Solution Package.
The team expects the solution to deliver measurable improvements across operations, customer experience, cost and energy. Early results on champion data point to up to a 60% reduction in mean time to detect, up to a 60% reduction in mean time to repair, fault identification up to 70% faster, and around 40% fewer incidents escalated to human operators. Together these translate into multi-million-euro annual savings at a single Tier-1 operator, through fewer truck rolls, lower energy bills, fewer SLA penalties and less overtime.
The approach also has a clear sustainability benefit. Smarter radio power management delivers 15 to 20% energy savings measured against real cell-level consumption, not modelled estimates. Because the radio accounts for most of a network's electricity use, this is one of the largest single levers operators have to cut Scope 2 emissions. Applied across the global mobile operator base, even partial adoption points to gigawatt-hours of electricity avoided each year, alongside better SLA conformance and higher customer satisfaction.
Agentic NOC shows how CSPs can use standards-based integration, intent-driven automation and agentic AI to progress toward Level 4 and beyond autonomous operations. Because the design is anchored in TM Forum standards and Open APIs, the components are portable across different operators rather than locked to a single vendor. Champions across North America (T-Mobile USA), EMEA (Turknet, Axian and Telefónica) and Asia-Pacific (Digital Nasional Berhad) show the approach holds up against real operator data in very different networks, with participants Nagarro, Enfec, Snowflake, DigitalRoute and Sutherland delivering the platform and agents.
For operators, the value lies in turning AI and automation into real operational outcomes: faster detection and repair, fewer incidents reaching people, lower energy use and more resilient customer experiences. The team offers the work as a production-ready blueprint that operators and vendors can carry forward beyond the show.
