LG Uplus AI agents and digital twins drive autonomous network strategy
LG Uplus is moving from rule-based automation to closed-loop autonomy, using AI agents and digital twins to accelerate toward a fully autonomous network by 2028.
Roadmap to a fully autonomous network by 2028
The Korean operator outlined a plan to apply AI across fault response, overload control, and service quality optimization, positioning autonomy as a step beyond today’s scripted workflows and dashboards. Its core platform, the AI Orchestration Nexus (AION), is already automating repetitive operations and has contributed to a reported 70% reduction in customer complaints about network quality—an early signal that the approach is translating into measurable outcomes. The company plans to showcase these capabilities at MWC Barcelona 2026, underscoring growing operator interest in operational AI as 5G matures and traffic patterns become more volatile.
How LG Uplus’s autonomous network works today
Specialized AI agents handle fault management through continuous monitoring, detecting weak signals of degradation, determining the scope and potential customer impact, and triggering remediation without waiting for a trouble ticket. Additional agents monitor service quality, correlate affected areas, and adjust configurations to stabilize experience metrics. To prevent congestion under fast-changing loads, agents proactively redistribute traffic and throttle hotspots before base stations become overloaded. A digital twin overlays the physical network with a virtual model of base station layouts and operational states, enabling on-screen inspection, scenario testing, and safer change management before pushing actions into the live environment.
Why autonomy matters for 5G-Advanced and enterprise SLAs
The timing aligns with the industry push toward 5G-Advanced and more demanding enterprise SLAs, where manual operations cannot meet scale, latency, or reliability expectations.
From automation to closed-loop autonomy
LG Uplus’s approach reflects the direction set by frameworks such as the TM Forum Autonomous Networks model and ETSI’s work on zero-touch network and service management, which emphasize policy-driven, intent-based, and closed-loop control. The goal is to shorten mean time to detect and repair, reduce OPEX tied to repetitive tasks, and improve compliance with service-level objectives—particularly where micro-outages and jitter can erode experience long before alarms fire.
Readiness for network slices, private 5G, and edge AI
As operators commercialize network slicing, private cellular, and edge workloads, they need automation that can interpret intent, guarantee quality under bursty conditions, and reconfigure resources in real time. AI agents and digital twins are building blocks for that future: they support deterministic behaviors at the cell and slice level, validate policies before deployment, and scale processes that would overwhelm human operators during events, disasters, or seasonal surges.
Strategic challenges to scaling autonomous networks
Scaling to autonomy introduces hard problems in data, governance, interoperability, and organizational change.
Data, observability, and MLOps for autonomy
AI agents are only as effective as the data they consume. Operators need unified telemetry spanning RAN, transport, core, and IT domains, normalized with low-latency pipelines and robust data quality controls. MLOps discipline is essential to handle model versioning, drift detection, explainability, and safe fallback strategies. Without strong observability and model governance, closed loops can magnify noise or behave unpredictably under rare conditions.
Multi-vendor interoperability and intent-based management
Real-world networks are heterogeneous, making multi-vendor control a gating factor. Adopting open interfaces—such as TM Forum Open APIs, ETSI zero-touch principles, 3GPP management specifications, and alignment with O-RAN for RAN policy and optimization—helps ensure AI agents can execute reliably across domains. Intent-based policies must be translated into vendor-specific actions with traceability, guardrails, and auditability to satisfy regulatory and security requirements.
Next steps: a practical roadmap to autonomous networks
Enterprises and operators can turn the autonomous network vision into concrete value by sequencing use cases, building the right foundations, and aligning with the ecosystem.
Prioritize high-impact closed-loop use cases
Start with domains where telemetry is rich and payback is clear: RAN congestion avoidance, anomaly detection in access and backhaul, proactive QoE stabilization for fixed wireless or video services, and automated fiber fault localization. Focus on loops that cut mean time to repair, reclaim capacity, or prevent customer-impacting events, then scale horizontally to adjacent functions.
Build digital twin and governance foundations
Develop a multi-domain digital twin that mirrors RAN, core, and transport topologies and traffic dynamics to test policies off-line. Establish model risk management, change control, and roll-back mechanisms before enabling fully autonomous actions. Define KPIs linking operational outcomes (e.g., MTTR, blocked calls, throughput variance) to business metrics (complaint rates, churn, SLA penalties) to prove ROI.
Align with standards and partner ecosystem
Use the TM Forum autonomous maturity model to baseline capabilities and chart progress. Leverage ETSI’s zero-touch concepts and 3GPP’s management standards for data models and lifecycle processes. Where RAN programmability is in scope, ensure compatibility with RIC-based optimization. Engage vendors to expose intent and telemetry APIs and to validate digital-twin compatibility early.
What to watch at MWC26 and beyond for autonomous networks
Proof points, ecosystem moves, and exportability will signal how quickly autonomy moves from pilots to mainstream operations.
Proof points, partnerships, and scalability
Key indicators include expansion from incident handling to predictive quality control and capacity optimization; the breadth of domains under closed-loop control; and evidence of safe autonomy at scale. Expect interest in how these capabilities integrate with OSS/BSS and whether the approach can be replicated across other markets. Any new alliances with equipment vendors, software suppliers, or cloud providers that accelerate data integration and model lifecycle management will matter.
Implications for vendors and enterprise buyers
Vendors will be pushed to deliver richer observability, intent APIs, and policy-driven control that work across mixed environments and map cleanly into digital twins. Enterprise buyers should see steadier performance, faster onboarding, and more transparent SLAs as operators industrialize these capabilities—particularly for latency-sensitive and mission-critical applications.
Bottom line: AI agents and twins accelerate autonomy
LG Uplus’s AI agent and digital twin program is an operational play with strategic consequences for 5G-Advanced and beyond.
Analyst perspective
By pairing domain-specific AI agents with a digital twin and an orchestration core, LG Uplus is moving decisively toward closed-loop autonomy, with early results that matter to customers. The hard work now is scaling across domains, vendors, and processes while hardening data, governance, and safety. For operators, the blueprint is clear: start with measurable closed loops, invest in observability and models you can trust, and anchor progress to open standards. Those who execute will translate autonomy into lower OPEX, tighter SLAs, and differentiated enterprise offerings as the industry enters the 5G-Advanced era.
