AI Transforming 5G RAN Operations: From Efficiency to the Path Toward Autonomous Networks

AI Transforming 5G RAN Operations: From Efficiency to the Path Toward Autonomous Networks

For modern telecommunications operators, applying management methods from the 4G era to 5G is no longer realistic. In addition to the realities of stagnant revenue, rising costs, and customer churn facing the industry, the burden on field operations—which must manage multiple network generations simultaneously—is greater than ever. Against this backdrop, a recent panel discussion featuring Rakuten Symphony brought together leaders from Axiata and Aira Technologies to discuss how AI is changing the economics and operations of 5G, and what is required to transition from initial efficiency gains to fully autonomous network management.

https://youtu.be/qK4yLQo7tX8
Video Description: AI is fundamentally changing how 5G networks are built, managed, and optimized. In this panel discussion hosted by The Network Media Group (NMG) at MWC26, leaders from Axiata, Aira Technologies, and Rakuten Symphony share practical insights on accelerating 5G RAN deployment, improving resource allocation, and leveraging AI to deliver superior user experiences in increasingly complex network environments.

AI: Opportunities Beyond Problem Solving
The discussion emphasized that AI should not be viewed merely as a stopgap measure for operational challenges. Rather, AI and automation can transform communication networks into programmable systems, allowing infrastructure to evolve into a system that flexibly adapts to changing conditions. Especially in markets like Southeast Asia, where prepaid contracts are the norm and customers can easily switch providers by swapping SIM cards, the ability to dynamically optimize network performance based on actual user behavior becomes a powerful competitive advantage. Operators that can leverage AI are not just technologically advanced; they are the ones capable of delivering true value in markets with demanding users and high performance expectations.

Areas Where AI Is Already Delivering Results and Future Outlook
Two areas were highlighted where AI is currently delivering the most visible results:
Energy Management:
Instead of relying on fixed schedules, machine learning-based traffic analysis allows for adaptation to real-world network phenomena, enabling extremely granular power control.
Root Cause Analysis:
Previously, operators could only analyze 1–3% of the data generated. By leveraging AI, this ratio can be increased to over 20–30%, dramatically improving analytical accuracy.
On the other hand, challenges remain for the autonomous, human-free operations that many operators aspire to. While the technical components are falling into place, trust in AI decision-making and the governance frameworks to guarantee that trust have not yet caught up. The speakers drew a parallel to autonomous vehicles, noting: 'For human operation to be replaced by electronic systems, being accurate is not enough. The ability to explain why a decision was made is an essential key for operators to fully trust AI.'

SMO: The Hidden Hero of Open RAN
Discussions about Open RAN often focus on the hardware atop the towers. However, this discussion argued that SMO (Service Management and Orchestration) is the most underrated key to transformation. SMO is the framework for centrally managing the entire communication network to enable automation and AI utilization. It liberates applications and analytical tools previously locked within specific vendor ecosystems, enabling third-party innovation. For operators with antennas and tools from multiple generations and vendors, SMO acts as an umbrella for integrated management, dramatically reducing the cost of introducing automation.

The Key to Success Lies in Building a Data Foundation
The discussion concluded by reiterating that data is the prerequisite for AI adoption. Currently, many operators have siloed systems for RAN, transport, CRM, OSS, and BSS. There is no common data foundation necessary for training AI models or performing correlation analysis. Without a data foundation, no matter how high-performance the AI is, it will not be useful. Furthermore, it was concluded that beyond the technical aspects, organizational change—abandoning operational habits built over decades and acquiring new skills—is essential. As Rishi Shukla of Rakuten Symphony noted, 'When a base station fails, surrounding Rakuten Mobile antennas automatically compensate for coverage and capacity to ensure customer experience is not compromised.' AI is already becoming a powerful weapon to prevent customer churn. Whether operators can truly leverage AI and realize autonomous networks in the future depends not just on technology adoption, but on building a data foundation and establishing organizational trust in AI decision-making.

Personal Observations and Reflections: How AI Is Changing the 5G Operations Field

The Complexity of 5G Operations: A Shift to a Different Game
I do not think it is accurate to describe the operational differences between 4G and 5G as an evolution. It should be called 'switching to a different game.' In the 4G era, base stations were integrated, and there was basically one vendor. Even when failures occurred, experienced engineers could handle most of it through intuition and experience. The number of items to manage was within the range that a human brain could grasp. In 5G, this premise collapses. Base stations are split into three layers—CU, DU, and RU—and in an Open RAN environment where each can be supplied by different manufacturers, the number of points to manage increases dramatically. Furthermore, network slicing requires running multiple virtual networks for different purposes on the same physical infrastructure simultaneously. You must maintain and optimize an ultra-low-latency network for factory production lines and a high-speed network for general consumers, each with different quality standards. This is not something humans can do manually. Moreover, in reality, 5G is being layered on top of 4G, which cannot be decommissioned. It is not that craftsmanship has become obsolete, but that the scale has reached a point where craftsmanship cannot be applied. Using AI is no longer a choice; it is a necessity.

AI Is Not a Rescue Plan for Laggards, but a Weapon to Make Leaders Stronger
The discussion pointed out that 'the biggest beneficiaries of AI are not the largest or most advanced operators, but those in the most competitive markets.' I believe this is correct, but there is another important aspect. AI is a technology where the difference in early investment snowballs: the operator that adopts it first accumulates data first, improves accuracy first, improves customer experience first, and reduces churn first. An operator that achieves AI-driven optimization early in the Southeast Asian prepaid market will have years of data accumulation by the time competitors catch up. If left unchecked, the gap in AI utilization will only widen, not shrink. I believe this is the type of investment where 'doing it eventually' will be too late.

The Unassuming SMO: Actually a Leading Presence
When it comes to Open RAN, hardware on cell towers gets all the attention because it is visible and easy to understand. However, because SMO (Service Management and Orchestration) is an invisible management layer, it is inevitably treated as a supporting role. There is a reversal here that is often overlooked. Even if you replace RUs and DUs with Open RAN-compliant ones, if the management mechanism remains dependent on a specific manufacturer, the cost of automation will not decrease significantly. Conversely, if you can centrally manage the entire network with SMO, you can layer automation and AI analysis on top, even if the RAN hardware is traditional. SMO is not an accessory to Open RAN; it is the gateway to AI utilization. In many cases, prioritizing the development of SMO over hardware renewal is more cost-rational. If you put this off, you create a structure where AI adoption will not yield results.

Explainable AI: The Barrier Is Not Technology, but Accountability
The biggest barriers to transitioning to autonomous networks were cited as 'trust' and 'the ability to explain why AI made a decision.' However, I think it is better to rephrase this more concretely. The real problem is that it has not been decided who is responsible when AI makes a wrong decision. When AI autonomously changes parameters and a large-scale failure occurs, is it the responsibility of the manufacturer that provided the AI, the operator that decided to implement it, or the engineer who supervised it? As long as these rules are not in place, field operators cannot press the approval button. The demand for 'explainable AI' is, before it is a technical problem, a problem of creating a framework to determine where responsibility lies. The example Rishi Shukla gave of 'surrounding antennas automatically compensating for coverage during a base station failure' is exactly the gateway to this question. When that automatic compensation causes another problem, who explains it and how? I believe that being able to provide an answer to this question is the real wall to realizing autonomous networks, rather than the maturity of the technology.

Whether You Can Build the Foundation Is the Turning Point
The fact that AI utilization without a data foundation is a pipe dream was repeated throughout the discussion. As long as RAN, CRM, and OSS exist in silos, AI can only make decisions based on fragmented information.

However, what concerns me is that data integration is always discussed as a 'technical prerequisite.' In reality, this is an organizational problem. For the team managing RAN and the team managing BSS to share the same data foundation, they must break down siloed organizational structures and the authority and processes that each department has protected. This is not a matter of technical investment, but a management decision.

Now that the complexity of 5G has made AI utilization inevitable, what is being questioned may be management's resolve rather than the choice of technology. I felt that this discussion was packed with universal challenges regarding AI adoption, not just issues specific to the telecommunications industry.


How AI Is Automating and Optimizing 5G RAN Deployments

https://symphony.rakuten.com/blog/how-ai-is-automating-and-optimizing-5g-ran-deployments

How AI Is Automating and Optimizing 5G RAN – Blog | Rakuten SymphonyLeaders from Axiata, Aira Technologies, and Rakuten Symphony share practical insights on how AI is transforming 5G RAN operations in this latest blog.