Level 4 autonomy starts with showing the value

Commercial context

Making Autonomous Networks Level 4 (AN L4) a working reality is one of the most ambitious goals in telecoms. Level 4 is a major milestone in TM Forum’s autonomy framework, where automated networks reason, act, and adapt independently to their objectives, with minimal human involvement.

Led by Huawei, Indosat Ooredoo Hutchison (IOH) Indonesia, Vodafone Turkey, China Mobile and Telenor, the 'Make AN Level 4 Real' Catalyst focuses on designing, deploying, and validating use cases that meet this standard. Rather than treating Level 4 as a future target, the team has built and tested live solutions in areas where autonomy has high value and clear operational payoff.

The team focused on three advanced use cases: cross-domain fault management, private line quality assurance, and RAN energy optimization. These were chosen for their complexity and potential to demonstrate both single-domain autonomy and cross-domain collaboration.

A key part of the challenge was to bridge today’s operations and true Level 4 capabilities. This meant tackling three interlinked challenges: understanding the current maturity of operations, designing agentic AI systems capable of true autonomy, and measuring their operational and financial impact. Together, these guided the team’s approach to operationalizing AN Level 4 in a measurable, replicable way.

The solution

The solution consists of an intelligent chain-of-thought (CoT) orchestration model. This model coordinates AI agents across network domains and allows them to reason through complex scenarios. Rather than simply execute commands, it analyzes the problem, generates a solution, and acts on it.

In the cross-domain fault management use case, for example, the CoT engine launches an agent to investigate a fault. That agent then queries the digital twin to identify the root cause. Based on the findings, the multi-agents assisted by LLM generate a corrective action plan. The CoT then initiates a closed-loop response using service keepalive atomic capability, activating the Cell Outage Detection and Compensation (CODC) process automatically.

The process is fully autonomous and transparent. Teams can trace each action, building trust in AI-driven operations. It also ensures a consistent and scalable response across domains, something that traditional manual systems can’t guarantee.

The Catalyst leverages a wide range of TM Forum assets to support its design and ensure interoperability. These include specifications such as IG1274M and IG1412 for AI agent design, the Autonomous Operations Maturity Model (GB1042), and value measurement frameworks including VOF and MAMA. The project also integrates TM Forum Open APIs to enable dynamic interaction between systems. Among those used are TMF642 for alarm management, TMF656 for service problem resolution, and TMF921 for intent-based control.

These APIs ensure that the solution can interact with existing OSS/BSS environments in a standardized and future-proof way. Just as importantly, the project linked these technical frameworks to business value. The team used TM Forum value models to measure each scenario and quantify time savings, workload reduction, and revenue impact.

Application and wider value

The project team is already seeing measurable results in a live environment. At IOH Indonesia, the system now resolves nearly 40% of alarms and incidents without human intervention. This shift marks a major change in how teams manage operations, especially in domains like RAN and IP-RAN.

In MBB fault handling, AI agents have improved productivity by 20%, while AI copilots supporting field teams have reduced communication time per work order. Tasks such as change request validation, which once took 20 minutes, are now completed within 10 minutes. Fault management engineers using self-service tools have also cut around 15 minutes per work order when working with the network operations center (NOC).

Luthfi Auzan, VP - Head of Operations Transformation & Analytics at IOH, reports that "we could achieve a 20% improvement in NOC efficiency and reduced mean time to repair by 15% through this Catalyst." These gains translate into financial business outcome with benchmark: annual savings of approximately USD 2 million through operational excellence. On top of that, higher service availability has helped prevent customer churn, securing an additional of minimum USD 0.5 million in retained revenue.

Following this deployment, the team will contribute solution packages and AN level evaluation tools to the TM Forum Autonomous Networks project. These contributions will help other CSPs assess their maturity, design agentic AI systems, and link automation efforts to clear business value.

AN Level 4 is no longer a hype but a reality. With agentic AI, orchestrated reasoning, and open standards, CSPs can now make a transition from automation to fully autonomous, and self-governing networks. When they do, the prize will be achieving competitive advantage, enhancing customer experience and driving operational efficiency.

The project won the Best Moonshot Catalyst – Autonomous Networks Trailblazer award at TM Forum’s Innovate Asia event in November 2025.