- Published on
How Telcos Build Autonomous Networks with Agentic AI
- Authors
- Name
- DeepHorus
How Telcos Build Autonomous Networks with Agentic AI
The telecommunications industry is undergoing a transformative shift as major carriers worldwide increasingly adopt agentic AI systems to build and manage autonomous networks capable of self-healing, self-optimizing, and self-configuring without human intervention. Unlike traditional AI models that respond passively to prompts, agentic AI operates as an autonomous digital worker β continuously monitoring, deciding, and executing tasks across sprawling network infrastructure. This evolution marks a pivotal moment for an industry grappling with exponentially growing data demands, complex 5G deployments, and shrinking operational budgets. By mid-2025, analysts estimate that over 70% of Tier 1 telecom operators have initiated agentic AI pilots across their network operations centers.
The push toward autonomous networks stems from a long-standing challenge in the telecom sector: the sheer scale and complexity of modern network infrastructure makes manual management increasingly untenable. A single major carrier like AT&T or Vodafone manages millions of network elements across dozens of countries, generating petabytes of performance data daily. Traditional network operations rely on teams of engineers responding to alarms and tickets β a reactive model that leads to prolonged outages, inconsistent service quality, and costly downtime. Industry estimates suggest that unplanned network outages cost global telecom providers over $40 billion annually, creating a compelling financial incentive for automation.
Several major telecom companies have emerged as frontrunners in deploying agentic AI for network autonomy. Deutsche Telekom announced in early 2025 that its autonomous network platform, powered by agentic AI agents developed in partnership with Microsoft Azure, now handles over 40% of routine network configuration tasks without human oversight. Similarly, SK Telecom in South Korea reported that its AI-driven network operations system reduced fault resolution times by 65% during Q1 2025 pilot tests. Meanwhile, BT Group in the United Kingdom has deployed agentic AI agents across its Openreach fiber network to predict and preemptively repair potential faults before they impact customers, achieving a reported 30% reduction in customer-reported service issues over a six-month period.
The technical architecture behind these agentic AI systems typically involves multiple specialized AI agents working in concert across different network domains. One agent mayθ΄θ΄£ radio access network optimization, continuously adjusting antenna parameters and power levels in real time, while another agent manages core network traffic routing and load balancing. These agents communicate with each other through orchestrated decision frameworks, allowing them to collaborate on complex scenarios such as regional traffic surges during major events or emergency situations. Nokia and Ericsson have both launched commercial agentic AI platforms in 2024-2025, offering pre-built agent frameworks tailored specifically for telecom use cases.
The implications of this shift extend far beyond operational efficiency. As networks become truly autonomous, the role of human network engineers is evolving from hands-on troubleshooting to strategic oversight and exception handling. Qualcomm's research division published findings in March 2025 indicating that agentic AI-driven networks could reduce operational expenditure by 25-35% over a five-year deployment horizon. However, this transition also raises significant concerns around accountability and governance β when an autonomous AI agent makes a decision that results in a service outage, questions of liability and responsibility become legally and ethically complex. Regulators in the European Union and the United States have begun drafting guidelines for AI autonomy in critical infrastructure.
Industry leaders have largely praised the progress while urging caution. Deutsche Telekom CTO Claudia Nemat stated at Mobile World Congress 2025 that "agentic AI represents the most significant leap in network management since the transition from 3G to 4G, but we must ensure robust human-in-the-loop safeguards remain in place." Ericsson CEO Borje Ekholm echoed these sentiments, emphasizing that autonomous networks must be designed with transparency and explainability as core principles. Meanwhile, advocacy groups such as the Open Networking Foundation have called for open standards to prevent vendor lock-in and ensure interoperability across different agentic AI platforms.
Looking ahead, the trajectory toward Level 4 and Level 5 autonomous networks β where AI agents can independently manage entire network ecosystems with minimal human intervention β appears inevitable. GSMA Intelligence projects that by 2028, over 60% of global telecom network operations will be managed by agentic AI systems. The convergence of agentic AI with digital twin technology, where virtual replicas of physical networks enable AI agents to simulate and test decisions before real-world deployment, is expected to accelerate this timeline. As the telecommunications industry continues its journey toward full autonomy, the companies that master agentic AI orchestration will gain a decisive competitive advantage in delivering faster, more reliable, and more cost-effective connectivity to billions of users worldwide.
