At a focused panel session titled “Intelligent Core Network Empowers High Stability Autonomous Networks”, industry experts from Huawei, TM Forum, and Telkomsel (TSEL) gathered to share progress, challenges, and innovative approaches in building next-generation autonomous core infrastructure.
The session highlighted a growing consensus: stability is the bedrock of autonomy in the core network. As Eric Luo, vice president of Huawei Core Network, put it, “Stability is the foundation, without it, efficiency is meaningless. I think it has already been a consensus among the industry.”
Huawei emphasised that achieving AN L4 in the core network, given its strategic and sensitive nature, demands a dual focus on high stability and high efficiency. While automation aims to reduce human intervention, Huawei insists that any advancement must first ensure service continuity, incident prevention, and proactive fault handling.
Huawei’s own path toward AN L4 leverages foundation model-driven AI technologies. These promise to boost automation in high-value scenarios, but challenges remain, including:
Model hallucination and accuracy gaps for operations and maintenance (O&M)
Limited self-learning and generalisation for complex issue resolution
High costs of improving data quality and reasoning engineering
Lack of industry-wide evaluation standards for AI model performance in network operations
Huawei called for stronger collaboration across the ecosystem, emphasising the need for shared practices and benchmarks.
For Telkomsel, the stakes are even higher. Operating across 17,000 islands in a region prone to natural disasters, Telkomsel sees stability as a survival imperative.
“High stability is not just a performance metric—it’s a survival requirement,” said Trihan Marsudi, GM of network digitalisation and analytic platform at Telkomsel. He explained how centralised fallback systems are often impractical in Indonesia’s geography, making decentralised, intelligent recovery essential.
“True autonomy must come with resilient-by-design architecture, where AI doesn’t just optimise performance but guarantees stability under worst-case scenarios.”
As CSPs strive to reach higher levels of autonomy, the TM Forum has been refining its Autonomous Network Level (ANL) evaluation matrix, particularly in light of growing activity in the core network domain.
Olta Vangjeli, programme director for cloud-native IT & networks at TM Forum, highlighted a major shift: “Intelligence without stability is dangerous.” The organisation is now working to better align AN evaluations with fault resilience, recovery speed, and service protection metrics.
However, as participation in the evaluations increases, evaluation inflation has emerged as a concern, with some operators allegedly overstating their level of automation.
To address this, TM Forum is proposing greater methodological consistency and governance, including:
Granular, scenario-based scoring
Stronger ties between automation and user experience assurance
Introduction of third-party auditing mechanisms to validate scores
Telkomsel supports this approach. Marsudi argued for a three-party co-evaluation mechanism involving the vendor’s tech transparency, the operator’s real-world use-case validation, and third-party review of outcomes.
Marsudi emphasized that KEIs, including metrics like call setup success rates during disasters and recovery times after signaling floods, are more than just operational indicators—they are critical benchmarks for assessing whether automation delivers real value to users.
One of the session’s most compelling contributions came from Telkomsel, which shared best practices and innovations aimed at predicting and preventing instability in its core.
In partnership with Huawei, Telkomsel is exploring a system called MDAF (Management Data Analytics Function), which enables Telkomsel to localise and contain faults faster than ever before.
Marsudi explained how the model works: “Based on multi-dimensional traffic modelling and anomaly behaviour learning, a signalling storm prediction model is developed. This allows us to pre-emptively identify high-risk zones before the storm occurs.”
Once a risk is detected, MDAF simulates potential pressure scenarios and recommends proactive measures like flow control, helping to maintain service even during sudden usage surges.
This marks a shift from reactive automation to proactive stability assurance: a hallmark of true Level 4 autonomy.
Throughout the discussion, a recurring theme was the need for cross-industry cooperation to standardise methodologies and promote meaningful progress.
Huawei urged stakeholders to work together on unified frameworks for evaluating stability , while Telkomsel echoed the importance of realistic, use-case-driven metrics.
Meanwhile, TM Forum says it is committed to strengthening its role as a neutral body guiding the evolution of autonomous network assessments.
The panel concluded with cautious optimism. While technical hurdles remain, particularly in the maturity of AI models and consistency of evaluations, the industry now shares a clearer roadmap: intelligence must be tied to stability, and autonomy must be anchored in accountability.
RELATED STORIES
