From cable cuts to continuity: how AI and intent drive real-time service recovery

Commercial context

Global telecom infrastructure is under constant strain. Outages, cable cuts, traffic spikes, and cross-domain complexity all erode service continuity and stability. These failures often result in manual responses that are slow, error-prone, and expensive. In the existing climate, maintaining service resilience with manual operations is no longer viable.

Tolerance for service interruptions is wearing thin among customers. Enterprise demand SLAs that hold during disruptions. CSPs face pressure to cut costs while scaling delivery across mobile, optical, edge, and cloud. Current systems can’t meet these demands. Operations are still fragmented, reactive, and difficult to scale.

'The AI-driven network automation for traffic and service resilience - Phase II' Catalyst reimagines service assurance and traffic resilience using AI, automation, and intent. The architecture enables CSPs to move from fragmented workflows to end-to-end closed-loop operations, ensuring networks can monitor, plan, and adapt in real time.

The solution

In phase II of the Catalyst, the project team brought together a broader group of partners to embed the foundations for level 4 and 5 autonomous networks. The solution combines TM Forum Open APIs and ODA, genAI, predictive AI, digital twins, a large language model (LLM), Camara APIs, 3GPP and MEF to create a programmable, self-healing, and scalable network operations framework.

CSPs begin by expressing business goals using natural language. For example, an enterprise customer requests a CSP provide ‘remote operation of IoT service,’ specifying time and location as the intent. AI interprets this and translates this into machine-readable instructions. The platform uses APIs such as TMF622 (Product Order), TMF641 (Service Order), and TMF652 (Resource Order) to provision services across the network stack.

Once services are live, the network continuously observes itself. A real-time digital twin simulates operating conditions and forecasts potential issues. AI ops monitor live telemetry and historical data, predicting failures or performance drops before they impact customers. When risks appear, the system takes action — scaling resources, rerouting traffic, and reconfiguring topologies.

The closed-loop architecture spans optical, IP, edge, and cloud domains. Agentic AI coordinates actions across business, service, and resource layers. For example, if an undersea cable fails, AI predicts the impact and shifts traffic to alternative links—often before users notice any disruption. This enables uninterrupted connectivity, even during complex multi-domain incidents.

The Catalyst also embeds energy optimization too. The platform identifies low-energy paths and powers down unused infrastructure. These green operations reduce emissions and lower OPEX without affecting quality.

What makes the approach scalable is its standards base. It aligns with TM Forum’s Open Digital Architecture and integrates more than 30 TMF assets, including TMF620 (Product Catalog) and TMF621 (Trouble Ticket), and TMF642 (Alarm Management). This project also uses ODA such as: IG1242 (Component Inventory), IG1271 (Connectivity-as-a-Service: A Paradigm Shift in Connectivity Delivery), IG1251 (Autonomous Networks – Reference Architecture), TR290 (Intent Common Model) and many more. The solution works across vendors and domains, enhancing solution flexibility through ODA components and TM Forum Open APIs

Key system components are already running on real platforms. Partners use Infonova/BeyondNow, Stablenet/Infosim and NetAware/Shabodi, and live conversational interfaces to showcase the solution’s applicability. This isn't a demo — it’s a working model ready for Phase III deployment.

Wider application and value

The impact of this Catalyst is both technical and strategic. In pilot runs, it improved traffic and service resilience by up to 80%. Time to market fell by 60%, and automated operations cut OPEX by 30%—figures confirmed by the participating CSPs. Moreover, it also results in improved energy management, cutting energy consumption by 20%. The calculation was conducted by researchers from our academic partner institution, working alongside the project team.

These gains apply across industries. Critical sectors — including healthcare, transport, finance, and public safety — all require intelligent, resilient networks. This Catalyst provides a blueprint to meet that need, enabling CSPs to deliver programmable, high-assurance connectivity in real time.

Fundamentally, the Catalysts provides a new opportunity and revenue stream for CSPs, and increasing their strategic value. With intent-based orchestration, CSPs can launch connectivity-as-a-service (CaaS) products with built-in QoS, SLA tracking, and latency guarantees. Enterprises no longer have to guess what the network will do. They can ask for what they need — and get it.

A true test of the architecture came during a global connectivity disruption. In early 2024, a subsea cable failure off the West African coast impacted multiple countries—one of an estimated 90 to 120 annual fiber or subsea cable cuts worldwide. The outage exposed just how brittle traditional operations remain. Routing traffic manually took hours. The economic impact was immediate. This Catalyst was designed with exactly this kind of event in mind — enabling real-time rerouting, fault prediction, and service continuity before customers even notice a problem.

As the project lead and Manager Connectivity & Internet Based Platform for Telin, Delfi Gunardy, explained: "this project is the result of bold collaboration between global leaders. CSPs include Telin, NTT, Orange, MTN, Verizon, Telecom Italia, Chunghwa Telecom, and Seacom. Technology partners include Infosim, Beyond Now, Shabodi, DGS, and Tallence AG, as well as academia such as the University of Würzburg. Their shared vision has produced a solution that is modular, standards-aligned, and operationally ready."

This Catalyst prepares a foundation for what network operations should become: predictive, adaptive, and intelligent. By turning intent into action and automation into assurance, this Catalyst shows what it takes to build a network that never waits and never stops.