A standards-based genAI platform for telecom service assurance

Commercial context

GenAI has arrived in telecom with big expectations, but few operators have turned it into structured, repeatable practice. For most CSPs, it still sits at the edge of operations: disconnected from real-time orchestration, not embedded into decision loops, and rarely aligned with business outcomes. Existing automation remains siloed, reactive, and rule-based — unable to handle dynamic network conditions or drive end-to-end service quality.

The 'Harnessing GenAI's business value for operations transformation - Phase III' has been set up to address this. Its objective is to embed genAI and agentic AI into core service operations. Not as pilots, but as coordinated, policy-governed agents that can scale across mobile, broadband, and customer experience domains. The goal is to improve performance, raise autonomy, and deliver measurable business value, all through a standards-based, modular framework built on TM Forum Open Digital Architecture.

The solution

Now in its third phase, the Catalyst concentrates on genAI autonomous capabilities. It continues to evolve from the Copilot model in Phase 2 to the Agentic AI autonomous model in Phase 3. This phase connects those pieces into a coordinated system — with agentic collaboration, digital twin simulation, and real-time value tracking. By embedding AI into live service domains and measuring its business impact, the project demonstrates operational intelligence at scale.

The team designed a modular, AI-native operations architecture which brings together genAI, digital twin networks, and autonomous agents in a closed-loop system. It covers real-time service assurance, experience management, and proactive network support.

The architecture is organized into three layers. At the top is the application layer, where genAI and agent-based co-pilots engages with use cases. The include mobile core service assurance, RAN keepalive monitoring, customer experience optimization, and home broadband support. The system is more than a dashboard, it acts on network signals, customer feedback, and performance metrics in real time.

In the middle is the agent layer, which uses agentic AI to coordinate decisions. These agents are not one-off models. They work together across domains, using frameworks like LangGraph and Model Contextual Protocols to maintain memory, context, and goal alignment. The co-pilot system applies this logic to service scenarios, helping operators respond to complex issues with minimal intervention.

At the foundation is the digital intelligent platform. It includes a large language model (LLM)-based genAI system and a digital twin network (DTN) that mirrors live conditions in the mobile transport layer. The DTN enables agents to simulate changes, predict outcomes, and resolve problems before they escalate. The digital intelligent platform forms a pivotal layer underpinning TM Forum’s AOMM (Autonomous Operations Maturity Model) and MAMA frameworks to track performance and maturity.

The project’s use of TM Forum assets is extensive. It draws on ODA definitions, eTOM business process models, SID for data modeling, and more than 30 supporting guidebooks, reference models, and APIs. This ensures the architecture is not only innovative, but standards-based and ready for adoption.

Wider application and value

The Catalyst demonstrates how genAI can deliver measurable improvements to service operations, with clear impact on customer experience and business results. One use case, focused on mobile traffic performance, cut data loss by 18.2%, saving over 1,000 TB per year. In another case, predictive assurance secured more than one million dollars in annual revenue by reducing churn, complaints, and network downtime.

As explained by Luthfi Auzan, VP - Head of Operation Transformation & Analytics at Indosat Ooredoo Hutchison (IOH): "Importantly, the architecture enables new services. For example, the project shows how 'latency-as-a-service' can be launched using a transport-layer digital twin. This offering alone has the potential to generate a few million dollars in new B2B revenue annually."

These results are mapped against TM Forum’s Value Operations Framework (VOF), showing how genAI and agentic AI raise autonomy levels, improve key performance indicators, and align operations with strategic goals. In particular, the project quantifies the shift from AOMM Level 3 to Level 4 maturity — where automation moves from reactive to proactive, and decisions shift from static rules to dynamic business logic.

This Catalyst provides a model for how to integrate genAI into everyday workflows and bridge the gap between intent and execution. Crucially, it elevates service operations with traceability and control. This gives teams the tools to move faster, with more confidence, and with clearer business impact. As CSPs push toward autonomy and efficiency, this project shows what it takes to turn genAI into a foundation, not a feature.