Managing a 5G network the way operators managed 4G is no longer viable. A recent session moderated by Rishi Shukla, APAC Sales Head at Rakuten Symphony, brought together leaders from Axiata and Aira Technologies to examine how AI is changing the economics and operations of 5G RAN and what it will take to move from early efficiency gains to fully autonomous network management. The discussion was grounded in the practical realities facing operators in APAC and beyond: flat revenues, rising spectrum costs, subscriber churn, and a workforce being asked to manage more network generations simultaneously than ever before.
Speakers:
Watch the full interview.
The panel pushed back on the familiar framing of AI as a response to operational pain. The more compelling argument, the leaders suggested, is the opportunity it creates. AI and automation bring programmability to the network and the ability to treat infrastructure as a fluid, adaptive system. In markets like Southeast Asia, where prepaid dominates and a subscriber can switch operators by buying a new SIM card, the ability to dynamically tailor network performance to real user behavior is a competitive necessity.
The panel was clear: the operators most positioned to benefit from AI are not necessarily the largest or the most technically advanced – they are the ones operating in the most sensitive, performance-dependent markets.
The panel was grounded on the question of where AI is actually producing results right now. Energy management was the clearest example: machine learning applied to traffic pattern analysis enables far more granular power management than static heuristics ever could. Instead of blanket schedules that power cells up and down on a fixed timer, AI-driven models adapt continuously to what is actually happening in the network.
Root cause analysis was the second area of measurable impact: operators today are analyzing perhaps 1 to 3 percent of the data their networks generate. AI makes it possible to work with 20 to 30 percent or more and the difference in diagnostic precision that produces is substantial.
Closed-loop automation, the panel acknowledged, is still ahead of where most operators are today. The technical elements are converging, but operator trust in AI-driven decisions (and the governance frameworks needed to support that trust) have not yet caught up. The leaders drew an instructive parallel: just as an autonomous vehicle requires a driver to be replaced by electronics, a fully autonomous network requires operators to trust an AI-based network driver to make and execute decisions without human intervention. Building that trust, the panel argued, requires more than accuracy. It requires explainability: the ability to show not just what decision the AI made, but why.
When operators think about Open RAN, the conversation typically starts with hardware on the tower. The panel argued this misses the more transformative contribution: the SMO. The SMO layer opens the door to third-party innovation at the network management level – applications and analytics that were previously locked inside vendor ecosystems. For operators carrying legacy infrastructure from multiple generations, this is significant: the SMO provides a single management umbrella capable of abstracting across traditional and open network elements alike, dramatically reducing the cost of introducing new automation capabilities.
The panel returned repeatedly to data as the precondition everything else depends on. Operators today typically have no unified data foundation: RAN, transport, CRM, OSS, and BSS systems operate in silos with no common pool against which AI models can be trained and correlated. Without that integration layer, the AI will not be smart.
The leaders also pointed to the human dimension: deploying AI requires organizations to unlearn deeply embedded operational habits built over decades, and to develop entirely new skill sets.
“Rakuten has developed AI-based analytics that run when any site is down, so that the periphery cluster or other sites can come up to help with the coverage and capacity without impacting the customer experience. This is a good way to stop customer churn.”
