Most industry applications of AI have focused on extracting insights from customer and network-generated data through big data and machine learning. Now, generative AI solutions like ChatGPT and Claude are enabling telecom companies to automate various processes and introduce new customer services more easily.
According to a recent AWS survey, AI adoption in telecoms is projected to reach 48 percent in the next two years. Many of the world’s largest telecom companies are already integrating these emerging technologies and dedicated language models into their operations. For instance, South Korea’s SK Telecom quickly positioned itself at the forefront of this trend by announcing, shortly after ChatGPT’s release, that AI would be central to its business strategy. Partnering with Deutsche Telekom, they developed a telecom-specific large language model (LLM) for efficient and quick deployment of generative AI models. These models, trained on billions of industry-specific parameters, are capable of handling complex tasks with less power and equipment.
Communications service providers can incorporate AI and LLMs into their daily operations in several transformative ways. In network optimisation, real-time analytics will continuously monitor every aspect of the network, allowing it to adapt to changing conditions seamlessly, self-optimise, and self-heal from faults. Network quality of service can be dynamically adjusted to reduce strain and energy usage during periods of lower demand. Autonomous drones equipped with picocells can supplement networks during peak times. AI can also optimise complex operations, improve service assurance, and lower operating costs by using real-time reports to monitor traffic and preemptively resolve potential issues. Analysis of behavioural data can further enhance services and predict usage trends. A model from the University of Surrey illustrates how an autonomous network using constrained combinatorial optimisation with deep reinforcement learning can save 76 percent in resources and reduce energy use, albeit with a 23 percent increase in compute costs. Digital twins will provide real-time insights into system performance, predict failures, and simulate scenarios to improve decision-making and efficiency in future network planning.
In business operations, AI will largely automate daily processes, helping telcos reduce operational costs, optimise supply chains, manage inventory, and identify new business opportunities. AI agents will manage billing and revenue operations, preventing errors and revenue leakage while ensuring correct customer charges. With most processes automated, human intervention will focus on areas where it adds the most value. Product creation will be streamlined through natural language commands, allowing new products and packages to be developed in seconds. Image recognition technology will enable auto-creation from sketches and drawings, revolutionising product conception and deployment. Products will be dynamically created at the point of sale, tailored to individual customer histories, business goals, and margins to maximise profits. Augmented reality will assist network management by enabling technicians to visualise and resolve complex issues remotely, reducing downtime.
Customer experience will also be transformed by AI. Customer profiles will be generated based on behavioural patterns, with chatbots capable of efficiently handling inquiries and providing responsive customer service. This will reduce wait times and improve overall satisfaction. Marketing efforts will be data-driven and precisely targeted to match customer preferences and behaviours. Virtual sales and support staff will communicate with customers through individually tailored agents, effectively acting as dedicated account managers with comprehensive access to customer interaction histories. This will allow for natural language communication in the customer’s preferred language and assistance with tasks like top-ups, renewals, and upgrades. Marketing campaigns will be automatically generated based on customer behaviour and preferences, offering hyper-personalised experiences and product recommendations. Emotion recognition and sentiment analysis will continuously monitor and adjust interactions. AI-generated insights will enable proactive assistance, predicting customer needs before issues arise.
Despite the advancements, it is unlikely that telecoms will become entirely AI-driven with no employees. Even the most advanced LLMs are prone to logical flaws, hallucinations, and bias. However, as these models improve and effective business use cases are identified, telecoms could operate with a leaner, more geographically distributed workforce. This does not necessarily mean mass layoffs. The telecom sector is a leader in workforce reskilling, with a significant percentage of employees retrained and companies like Telefónica running Europe’s largest internal retraining program. AI can complement human efforts by automating routine tasks and managing virtualised networks while fostering a culture of human-driven collaboration. Reducing headcount should be approached cautiously, as replacing customer-facing roles entirely with AI could negatively impact customer experience. Surveys indicate that most customers prefer human interaction and can be frustrated by the inability to speak to a real person. Nonetheless, well-implemented AI chatbots have shown to achieve similar satisfaction levels more efficiently than traditional services.
While AI is expected to reduce the number of jobs in telecoms, many new roles will emerge, although a net reduction of 14 million jobs across all sectors is anticipated in the next five years. As ongoing legal cases against generative AI providers are resolved, it will take time before these technologies significantly impact the industry beyond the current hype. The human element remains essential for guiding AI development and integration. There is a significant opportunity for vendors to implement AI-powered solutions tailored to their organisational data. Presently, only a minority of telecoms are building in-house models, with most opting to use off-the-shelf solutions that leverage proprietary data. This cautious approach is likely wise given the rapid evolution of generalist LLMs outperforming industry-specific models.
Telecom companies embracing AI and LLMs could see substantial improvements in customer satisfaction and operational efficiency, with enhanced predictive capabilities and security. While the all-AI telecom remains a distant prospect, the rapid pace of technological change suggests its realisation may come sooner than expected, surprising those unprepared.
Cerillion plc (LON:CER) is a leading provider of billing, charging and customer management systems with more than 20 years’ experience delivering its solutions across a broad range of industries including the telecommunications, finance, utilities and transportation sectors.
