From reactive operations to customer-aware autonomous networks

From reactive operations to customer-aware autonomous networks

Telecom networks are becoming more complex, while customers are becoming less tolerant of poor service. Traditional network operations center and service operations center models were designed for a simpler environment, where teams could detect incidents, open tickets, and resolve problems after they were identified. But in today’s networks, customers may experience the impact before operations teams see the issue. By the time a ticket is opened, customer experience and net promoter score may already have suffered.

The Catalyst project, CX optimization via AI-driven SOC over autonomous networks - Phase II, addresses this challenge by connecting automation directly to what customers are experiencing. Building on a live Tier-1 production deployment from Phase I, championed by Claro Colombia, Claro Brasil and Libyana Mobile, Phase II expands the architecture into additional network domains and use cases, moving closer to Autonomous Networks Level 4. The result is intended to be a reusable, open, and standards-based blueprint that operators can adapt to their own environments, regardless of vendor mix or network maturity.

The challenge: automation has to understand customer impact

Most CSPs have already automated parts of their operations, but many automation workflows still remain technical, reactive, and disconnected from real customer impact. Network alarms, performance data, service tickets, and customer experience signals often sit in separate systems. That makes it difficult for operations teams to understand which incidents matter most, which customers are affected, and which remediation action will deliver the best business outcome.

The project team argues that the question is no longer whether to automate, but whether automation is fast enough, smart enough, and connected directly enough to the customer experience. If automation is driven only by technical thresholds, it can miss the issues that customers feel most acutely. If it lacks the ability to simulate or validate decisions, it can also create new operational risks. Operators need closed-loop automation that can interpret intent, learn from prior decisions, assess likely outcomes, and act safely across complex brownfield environments.

Embedding AI agents across the closed-loop lifecycle

The Catalyst embeds AI agents across the full closed-loop automation lifecycle. In the first layer, intent-based AI defines closed-loop automation strategies and continuously evaluates how well they are working. It draws on insights from a knowledge engine and simulations in digital twins, helping automation logic evolve as network conditions, services and customer expectations change.

In the second layer, AI agents participate directly in decision-making and execution. When live network events occur, the agents formulate remediation plans, validate those plans through digital twin simulations, and refine actions before they are safely deployed. Successful strategies are then fed back into the knowledge engine, creating a continuous learning cycle that improves future decisions.

This approach moves the SOC from a reactive function toward a customer-aware, AI-driven operating model. Rather than simply responding to alarms, the solution is designed to identifyservice degradation, understand customer and business impact, select the right remediation path, and execute corrective action through closed-loop automation. Because the architecture is modular and vendor-agnostic, the team positions it as a practical reference model for operators that need to modernize operations without replacing their existing technology stack.

Built for brownfield scale

One of the most important aspects of the Catalyst is that it is not presented as a lab-only demonstration. Phase II builds on production experience and is designed for brownfield environments where operators must work across existing systems, fragmented data sources and mixed vendor domains. The team’s goal is to provide a reference architecture that can be replicated by different operators and adapted to different levels of network maturity.

That matters because the path to Autonomous Networks Level 4 will not be achieved through isolated use cases alone. CSPs need reusable patterns for intent, decisioning, simulation, execution, and learning. By validating the architecture against TM Forum’s Value Operations Framework and aligning it with Open APIs and autonomous network principles, the Catalyst helps create a common foundation for AI-driven service operations that can scale across domains.

TM Forum assets used

The Catalyst uses a set of TM Forum assets to support interoperability, closed-loop automation, and standards-based adoption. These include TMF639 Resource Inventory, TMF642 Alarm Management, TMF628 Performance Management, TMF753 Closed Loop Management, TMF702 Resource Topology, TMF921 Intent Management and TMF640 Service Activation and Configuration.

Together, these assets provide a standards-aligned foundation for inventory, topology, alarm handling, performance insight, intent management, service activation, and closed-loop control. This helps operators avoid point-solution automation and instead build toward a more consistent, reusable and interoperable operating model for AI-driven autonomous networks.

Business impact and industry benefits

The project team expects the solution to deliver measurable improvements in cost, speed, operational efficiency, and customer experience. The stated business impact is to reduce OPEX by 25% to 35% and cut mean time to repair by 50% to 70%, moving resolution times from hours to minutes. The team also expects 60% to 80% of routine incidents to be resolved autonomously, reducing manual NOC effort by up to 60%.

For customer experience, the potential impact is equally significant. By detecting anomalies through real-time, user-experience-aware data and resolving issues before they escalate, the project aims to reduce customer trouble tickets by 30% to 40% and drive measurable NPS improvements in high-ARPU segments within 90 days. For customers, that means fewer disruptions and more reliable digital experiences. For operators, it means a clearer link between network automation, customer loyalty, and commercial performance.

Felix Reuben, Principal Engineer Technology Strategy at Safaricom, said the growth of AI and the demand for seamless customer experience mean telcos and ISPs must transition from basic connectivity providers to technology providers. He added that customer experience and operational efficiency are key enablers in this transition, and that the project “not only enables Telcos to enhance customer experience but also provides an opportunity for automated operational efficiency that will lead to improved service delivery and experience with an opportunity to scale to more use cases.”

Why it matters

This Catalyst connects two priorities that are often treated separately: autonomous operations and customer experience. As networks become more dynamic and AI-driven, CSPs need operations models that do more than reduce manual effort. They need automation that understands which actions will protect service quality, reduce customer friction, and support business outcomes.

By combining AI agents, digital twins, a knowledge engine and TM Forum standards, CX optimization via AI-driven SOC over autonomous networks - Phase II offers a practical route toward customer-aware autonomy. It gives operators a way to move from fragmented, reactive service operations to closed-loop decisioning that can sense, simulate, act, and learn. If replicated at scale, that could help the industry accelerate the journey to Autonomous Networks Level 4 while making the benefits visible where they matter most: in the experience customers receive.