Across APAC, autonomous networks powered by agentic AI, machine learning, and automation are enabling goal-directed decisioning, streamlining processes, and reducing operational cost. In a session moderated by Abe Nejad of The Network Media Group (NMG), industry leaders recently examined real-world strategies, challenges, and innovations driving self-optimizing, AI-driven operations.
Speakers:
Watch the full interview.
The session underscored the shift from classical AI to agentic AI – moving beyond predictive models and scripted workflows to systems that pursue intents, reason over context, and act. Classic ML continues to power planning, spectrum optimization, fault correlation, and energy savings. Generative AI accelerates access to insights through natural language. Agentic AI wraps these capabilities to identify repeatable patterns, trigger end-to-end automation, and close the loop with measurable outcomes.
Leaders emphasized an outcome-back approach to autonomy. Start with the decisions that matter (capacity design, proactive fault prevention, experience KPIs), then orchestrate data, policy, and automation to execute in real time. Practical impacts include faster MTTR via multi-domain root-cause analysis, dynamic resource scaling around predictable surges, and policy-safe closed loops that act before customers notice degradation.
The conversation highlighted that autonomy will remain human-in-the-loop for policy, security, and regulatory controls, while machines handle speed-critical decisions; the challenge is placing humans strategically so governance doesn’t stall automation. ROI signals include time-to-action, decision quality/precision, and pre-/post-impact on KPIs such as latency, availability, energy, and NPS.
Data remains the hardest problem. Telcos must improve quality, semantics, lineage, and access control to feed agents reliably, while avoiding new silos. The panel also flagged multi-agent synchronization as a coming differentiator: the ability to coordinate specialized agents safely across domains will separate pilots from production at scale.
Looking ahead, the stack itself is set to change. The leaders pointed to agent-native architectures – reimagining OSS and operations software so intent, policy, and closed-loop execution are foundational, not bolted on. With compute costs falling and capabilities compounding, the next 24 months were framed as a window to move from experiments to durable operating gains.
“The way the tech stacks, the networks, and the infrastructure are built today is going to be reimagined. Traditional software products designed to manage networks will transform into something completely different – using agentic AI as the native way to build those stacks. That’s where the real change begins.”
