The industry must focus on four key imperatives: reducing costs, driving growth, differentiating customer experience and ensuring secure, reliable operations. Reduce cost to serve Despite efforts to bring them down, CSPs’ operational costs remain stubbornly high, at 65- 70% of revenue; network operations alone will consume 50% of total operating expenses (OpEx) by 2027.8 Traditional AI has long contributed to cost reduction and efficiency, such as through predictive maintenance, with nearly two-thirds of AI professionals across CSPs and hardware or software providers reporting cost savings from AI use cases.9 GenAI builds on this by processing unstructured inputs and generating content, unlocking cost savings in two ways: Firstly, genAI enhances efficiencies through data democratization and automation of repetitive, structured tasks, such as network planning applications or generating customer emails. Natural language capabilities enable the enterprise- wide sharing of data and knowledge to improve decision-making, efficiency and outcomes. Secondly, CSPs can enable increased process automation across IT and network management, helping address persistent challenges and drive efficiencies. Across information technology (IT), CSPs report a high technical debt, which consumed 56% of IT spend in 2023.10 This debt results from piecemeal technology adoption, leading to fragmented systems, siloed data and legacy architectures that complicate modernization efforts. Only 7% of CSPs report being “fully satisfied” with recent modernization attempts.11 However, genAI’s ability to ingest unstructured data and produce blueprints, test scripts and standardized code components enables the automation of end-to-end software and data delivery life cycles. This comprehensive approach to reducing technical debt improves the speed, consistency and quality of modernization efforts while embedding security throughout. The transition of network infrastructure from legacy monolithic architecture to disaggregated hardware and software layers allows AI to play a bigger role in network automation. For example, together, traditional AI and genAI can create cost- effective network designs, automate multi-vendor component integration and continuously monitor performance to detect anomalies. They can also perform real-time root cause analysis within operations and ensure security configurations meet baseline requirements. When faults are detected, AI can propose solutions or autonomously adjust traffic routes to minimize downtime, enhancing service quality and continually progressing towards fully automated network operations. A human-on-the-loop12 approach is vital, ensuring efficiency and reliability while mitigating risks such as configuration drift (whereby standard configurations deviate over time) and model hallucinations (the creation of nonsensical or inaccurate outputs) through consistent human oversight. Drive business growth As CSP revenues are squeezed and enterprise value declines, the industry is exploring new growth opportunities. While only 27% of AI professionals across CSPs, hardware and software providers cite “meeting revenue targets” as a primary goal for AI use, 67% have reported revenue uplift in specific business areas attributable to AI.13 In business-to-consumer (B2C), AI is reshaping customer acquisition and retention strategies. GenAI enables the creation of highly personalized marketing campaigns and customer journeys powered by predictive models that anticipate individual behaviours. These capabilities allow CSPs to identify purchase propensity or churn risk, tailor offerings through granular customer segmentation, optimize bundling of traditional communication and adjacent digital products, and price them more strategically. In business-to-business (B2B), CSPs are increasingly moving up the value chain by capitalizing on their infrastructure provision and AI capability development. As existing infrastructure providers, CSPs can employ their data centres to offer innovative services across the full spectrum of the cloud, edge compute and connectivity. This includes hosting AI infrastructure with graphics processing units (GPUs) and inferencing capabilities to meet local enterprises’ needs and providing ICT (information and communications technology) services and off-the-shelf edge industry solutions in partnership with ecosystem players. This aligns with the growth in demand for AI computing power and cloud migration, which have driven a 35% growth in data centre spending, which stood at $318 billion in 2024 and is forecast to grow another 15.5% ($49 billion) in 2025 (versus 4.4% for communications).14 The digital infrastructure offering can position CSPs to support at least one of the key sovereign AI solution pillars,15 with some of the CSPs particularly well-positioned to do so, having developed their own sovereign large language models (LLMs).1.2 Industry imperatives Traditional AI and genAI can create cost effective network designs, automate multi-vendor component integration and continuously monitor performance to detect anomalies. Artificial Intelligence in Telecommunications 7

Artificial Intelligence in Telecommunications 2025