Autonomous network decisions start and end with dataAutonomous network decisions start and end with data

Telco profitability is challenging due to rising operating costs, competition from new services and entrants and sluggish revenue growth. Global annual revenue growth remains flat (0% between 2019 and 2023), according to Omdia's Global Telecoms Capex Tracker – 2023 (July 2024). To mitigate some of these issues, operators must prioritize digital transformation and automation to support new services. For example, Orange aims to reduce its capex by $646 million in 2025 and achieve highly autonomous networks. Telefónica and China Mobile have also committed to achieving high autonomy (TM Forum Level 4 shown in the figure below) by the same year.

Autonomy is essential to operators that want to achieve operational efficiency and savings, provide an enhanced customer experience and offer insight-driven services. The goal — autonomous networks leveraging AI technologies to sense, think and act — offers great potential but poses initial challenges. A long-term plan may be fully autonomous, self-orchestration, healing and optimizing zero-touch operations. However, operators must start the process, beginning with a network data strategy for many.

The foundations

Network decisions rely on data. If the data is flawed, biased or of poor quality, the outcome will be the same — garbage in, garbage out. According to TM Forum, data preparation accounted for around 80% of AI solution creation in 2023. Data precision is complex but central to improving customer experience, meeting operational and sustainability targets and retaining service quality. Operators must consider key data challenges as they integrate automation into their networks:

The journey to autonomy

Network autonomy begins with automating repetitive, low risk tasks and transitioning to AI-driven reactive operations. TM Forum classifies the autonomous levels (shown in the diagram) in six steps. Many operators are in the early stages (Levels 1–2), formulating strategies, technology approaches and organizational cultural changes. Full autonomy (Level 5) uses all forms of AI, such as predictive AI, generative AI (GenAI), large language models (LLMs), etc., to fulfill intent-driven business requests.

Predictive AI can enhance basic analytic capabilities by offering recommendations to optimize coverage and capacity based on user activity. It can predict churn by correlating customer experience metrics and helpdesk calls, anomaly detection from historical pattern learning, etc. AI models are faster at associating data, finding anomalies and building connections than humans, as they learn iteratively from past network conditions.

GenAI tools equipped with LLM interfaces address challenges such as long knowledge acquisition timescales, configuration generation, comprehension of complex queries for root cause analysis, etc. Some AI data platforms incorporate telco-specific model training, but generic models can require additional integration and training to ensure telco domain accuracy.

AI development platforms offer innovation hubs to co-develop and expand AI-based tools and apps, foundations of new services or efficiency measures. To simulate requirements for service introduction, digital twins can replicate the live network and provide safe environments to monitor unexpected behavior or anomalies.

Autonomous network levels

Reaping the benefits

Early automation use cases are starting to demonstrate significant performance and opex savings. For example, China Mobile has been an early adopter, integrating a GenAI solution to assist engineers with fault resolution and data analysis processing and reports. Using natural language dialogs (leveraging LLMs), engineers can query and retrieve information from manuals, reports and libraries to resolve faults. The GenAI solution also assists developers in data development and the creation of data analysis reports.

According to China Mobile, its network autonomy solution resulted in an 80% reduction in knowledge acquisition time and a 72% increase in data analysis efficiency over a one-year timeframe. Automation has also reduced China Mobile's network operating costs by around $7 million yearly.

Challenging market conditions are driving operators to integrate network automation. To reach full network autonomy, operators must transform network data, requiring stakeholder buy-in, new data governance frameworks, consolidation and AI data platforms. Change will evolve in stages, and operators must leverage predictive AI and GenAI capabilities across all layers, domains and services to support their end goal: full autonomy.

This blog is sponsored by Nokia.