Mobile operators must speed up the transition toward intent-driven network operations as the telecom industry faces rising operational complexity and cost pressures, according to a new report from GSMA Intelligence. The study highlights that achieving higher levels of network autonomy has become a strategic priority for telecom companies in 2026.
The above statistics prepared by GSMA Intelligence indicate Capex and ARPU trends in the telecom services industry during the last several years.
Network Autonomy Becomes an Industry Imperative
Telecom operators have long pursued automation to improve network efficiency and reduce operational expenditure (opex). However, most automation initiatives have been implemented in isolated silos across different network domains, limiting their overall impact.
According to the report, incremental automation alone will not deliver the cost reductions operators are seeking. Instead, telecom companies must move toward cross-domain automation that connects network functions and operations. Reducing operational complexity, improving predictability in costs, and achieving measurable opex savings are emerging as key investment drivers.
In this environment, shifting toward intent-driven operations – where high-level business objectives automatically translate into network actions – is becoming essential for telecom operators.
Building Blocks for Autonomous Networks
The report stresses that a foundational requirement for autonomous networks is a robust layer of AI-ready network data. Many operators have already centralized network data into large data lakes, but data streams remain fragmented across different systems and domains.
In addition, much of the available network data is not yet suitable for AI-driven analytics. Operators need to restructure and prepare data to support machine learning models, automation tools and AI agents that will power next-generation network operations.
Telecom companies are increasingly evaluating digital twin technology to simulate network environments. Digital twins allow operators to test complex scenarios where AI agents interact, negotiate and execute network tasks based on higher-level intents before deployment in live networks.
Agentic AI Emerging as a Key Enabler
Agentic AI is beginning to gain attention within telecom operations, although the technology is still in its early stages of adoption. AI agents are expected to act as a bridge between business intent and network execution.
These agents could convert strategic business objectives – such as improving network performance or customer experience – into executable instructions across telecom infrastructure.
However, for AI agents to function effectively in telecom environments, they must rely on deep telecom domain knowledge. As a result, several operators are investing in or developing telco-specific foundation models that can understand network terminology, processes and architectures.
Most Operators Still at Early Autonomy Levels
Despite ongoing innovation, most telecom operators remain in the early stages of network autonomy. According to the report, more than two thirds of operators have not progressed beyond Level 2 autonomy, which represents partial automation within specific domains.
At this level, network operations remain largely reactive and incident-driven, even though some automation exists.
There has been some progress. In 2025, a small group of operators demonstrated Level 4 autonomy in proof-of-concept environments and began investing in scaling those capabilities. However, most operators still expect to achieve Level 4 autonomy around 2030.
Level 4 autonomy represents a major shift toward proactive operations and declarative, intent-driven network management.
AI Agents Will Power Future Autonomous Networks
The report describes AI agents as the operational “foot soldiers” of autonomous telecom networks. These agents will translate high-level business objectives into real-time actions across network infrastructure.
To support this transformation, operators must develop a clearly defined network ontology that enables AI systems to interpret network relationships, processes and operational data.
Agents must also be deeply embedded within telecom operations rather than added as external automation tools. Without integration into telco-specific models and systems, AI agents may struggle to operate effectively across both modern cloud-native infrastructure and legacy network environments.
Early Use Cases for Agentic AI in Telecom
Telecom operators are already exploring several practical use cases for agentic AI. Key areas under evaluation include:
Automated fault detection and resolution
Incident reporting and management
Root cause analysis of network failures
Customer complaint resolution
Network optimisation and performance management
Customer experience and network management are emerging as the most promising early applications.
Data Quality Remains a Critical Challenge
Although telecom networks generate massive volumes of data and telemetry, much of this information is not yet structured or standardized for AI applications.
The report emphasizes that improving data quality is essential for achieving real network autonomy. Operators must focus on cleaning and consolidating network inventories, operational data and telemetry streams.
A horizontal data model that integrates data across multiple network domains is also required. This approach allows network insights to be analyzed centrally and converted into automated actions.
In addition, the data layer must be easily accessible for AI platforms and telco-specific models that rely on continuous data ingestion. Strong governance policies are needed to ensure data integrity and security.
Start Small With Autonomous Network Use Cases
The report recommends that operators adopt a use-case-driven approach to implementing autonomous network technologies rather than attempting large-scale transformations immediately.
Telecom companies should begin with targeted deployments of AI agents in specific operational scenarios. Once a particular use case proves successful, the solution can then be expanded across the network and combined with additional use cases.
Digital twins will play an important role in this process by enabling operators to simulate network behavior, test automation scenarios and train AI models before full-scale deployment.
Autonomous networks
Autonomous networks gained significant attention among telecom operators in 2025, with several leading companies advancing proof-of-concept deployments and trials to validate Level 4 network autonomy. According to industry developments tracked by GSMA Intelligence, multiple technology vendors, operators and industry groups are working to build the foundation required for large-scale autonomous network adoption.
Several major initiatives and partnerships were announced during the year to accelerate the transition toward AI-driven telecom operations.
In February, Nokia introduced new agentic AI capabilities designed to strengthen its autonomous network portfolio. The company aims to help operators automate network management and operational processes using AI agents.
In March, Deutsche Telekom partnered with Google to develop RAN Guardian, an AI-powered network agent focused on radio access network operations. The same month, Amdocs expanded its AmAIz platform with new network agents developed with support from Nvidia and Amazon Web Services to assist operators with automated network planning and deployment.
Industry collaboration also intensified in June when TM Forum, Huawei, Telefónica and Orange launched the “L4 is ON” initiative, aimed at accelerating the telecom sector’s progress toward Level 4 autonomous networks.
In September, China Mobile and Huawei received the TM Forum Excellence Award for their ‘Dark NOC’ autonomous network operations center, which operates with minimal human intervention and extensive cross-domain automation.
Progress toward Level 4 autonomy was also reported by operators. Advanced Info Service (AIS Thailand) announced in October that several operational domains in its network had achieved Level 4 autonomy, supported by AI agents designed to reduce downtime and optimize energy efficiency across its nationwide 5G infrastructure.
The same month, SoftBank launched a large telecom model (LTM) based on its in-house large language model called Sarashima to train AI systems tailored for the Japanese telecom environment.
Meanwhile, Digital Nasional Berhad in Malaysia and Ericsson achieved the world’s first Level 4 autonomous network certification from TM Forum.
These developments highlight the telecom industry’s growing commitment to autonomous network technologies, even as most operators continue working toward broader Level 4 adoption later in the decade.
As telecom networks grow more complex with the expansion of 5G, cloud-native infrastructure and AI-driven services, the shift toward intent-driven autonomous operations is expected to become a central pillar of the telecom industry’s next phase of transformation.
BABURAJAN KIZHAKEDATH
