Anatomy of an AI telco

Telcos are increasingly looking to AI and LLM-powered solutions to improve their current offerings and implement new, cutting-edge services. What will the AI-focused telco of the not-too-distant future look like?

The AI telco is coming sooner than you think.

To date, most industry use cases for AI have been limited to extracting insights from customer and network-generated data using big data tools and data scientists. Now, GenAI solutions, such as ChatGPT and Claude, are increasingly easy for telcos to leverage, using large language models (LLMs) to automate many different processes.

A recent survey by AWS found that adoption of AI in telecoms will hit 48% in the next two years, but many of the world’s largest telcos are already well on their way to integrating emerging technologies into their business processes.

South Korea’s SK Telecom was well ahead of the curve, when it announced mere weeks after ChatGPT’s initial release that it was putting AI at the core of its business, committing to raise investments in AI from 12% to 33% by 2028, with a focus on enterprise data, private networks and personalised content.

Meanwhile, SK Telecom partnered with Deutsche Telekom to develop their own telecom focused LLM for CSPs to deploy GenAI models efficiently and quickly. With their training parameters distilled to a few billion, these models can handle complex, industry-specific tasks, while requiring less power and equipment.

How can the telecoms industry incorporate AI, LLMs and chatbots into their everyday work processes? Here are three key areas we think could be radically transformed:

Network Optimisation

Combining AI with networks to automate processes and reduce the need for human intervention has been a dream of vendors for many years. As vast quantities of data are created, AI can be used to optimise complex network operations, improve service assurance and reduce operating costs.

• Software-defined networks self-build, self-optimise and self-heal, adapting to changing conditions seamlessly, dynamically adjusting QoS for different services based on demand.
• Machine learning continually optimises network performance, allocating resources efficiently and adapting to traffic patterns instantly, using real-time data to predict and resolve potential issues.
• Real-time analytics constantly check every area of the business and correct issues, shutting down quiet cells or reducing compute to save money. Traffic is monitored automatically, supplemented by autonomous drones carrying picocells to provide instant capacity upgrades when traffic increases or is predicted to increase.
• Digital twins for network planning, such as the NVIDIA Omniverse, provide real-time insights into system performance, predict potential failures and simulate various scenarios to enhance decision-making processes and improve overall efficiency.
• Strain on physical networks and energy usage from extra processing is optimised, leading to more environmentally sustainable infrastructure.

We’re already seeing the seeds of such networks; a new AI model developed by the University of Surrey, utilising “constrained combinatorial optimisation with deep reinforcement learning” can save up to 76% in network resources and decrease energy use, with only a 23% increase in computational costs.

Business Operations

To meet new workplace demands, whether that be a workforce that is older, leaner or more geographically distributed, AI will reduce operational costs, optimise supply chains, improve inventory management, and make suggestions for new business opportunities.

• AI-driven insights analyse customer behaviour and identify service improvements, enabling better decision-making, network optimisation and to predict customer trends.
• All BSS processes run automatically; any human interaction is purely through voice or gesture, with AI interpreting what is needed and designing based on that.
• Products are uniquely created for the individual at sale time, based on history, business goals and margins to maximise profits while keeping the customer happy.
• Natural language text or voice is used to build products in a matter of seconds; image-recognition can be used to auto-create new products or workflows from sketches and drawings.
• AI-powered augmented reality enables technicians to visualise and solve complex network issues remotely, leading to faster resolution times and reduced downtime.

By automating product creation, innovation and routine business processes, AI can oversee billing and revenue management operations, identifying errors and preventing revenue leakage, whilst ensuring that customers are charged correctly.

Customer Experience

To the AI telco, chatbots capable of understanding and resolving enquiries will help provide highly efficient and responsive customer service, reducing wait times and improving overall satisfaction, while back-end applications will be used to anticipate and proactively work against churn and towards acquisition and retention.

• Customer service is fully personalised using individually tailored avatars for each customer, with each having instant access to every interaction you’ve ever had with them.
• Data-driven and personalised marketing, based on analysis of customer behaviour and preferences for better product and package targeting.
• AI chatbots respond to customer queries in natural language and in any language, helping with tasks such as top-ups, renewals and upgrades. Emotion recognition technology gauges sentiment, enabling telcos to automatically tailor interactions and services accordingly.
• AI generates hyper-personalised plans and device recommendations based on usage and intent data, building consolidated customer profiles and identifying patterns of behaviour.
• Predictive insights into future customer behaviour will anticipate needs, prevent issues and offer proactive assistance before customers even realise they require it.

In the AI telco, all marketing efforts are data-driven, tailored to individual preferences and behaviours for precise targeting. Through AI-generated insights, hyper-personalised plans and device recommendations are crafted, leveraging usage data and intent analysis to build comprehensive customer profiles and anticipate future needs, enabling proactive assistance before issues arise.

How soon could we see an all-AI telco, with no employees? It’s unlikely for now – even the best LLMs are still prone to logical shortcomings, hallucinations and bias. In time though, telcos are likely to not need quite as many people to run them; while many new jobs will be created, many more are expected to be lost, for a net reduction of 14 million jobs across all sectors in the next five years.

However, just because you can doesn’t mean you should.

AI could replace much of the customer-facing side of a telco business, but doing so might seriously impinge on the experience. In fact, for all the possibilities of automating customer support, it’s worth remembering that 34% of consumers consider AI-generated material inappropriate for customer-facing content, with another study by Ipsos reporting that 88% of people prefer talking to a human. Nevertheless, when done well, AI chatbots have been shown to achieve similar customer satisfaction scores whilst being quicker than traditional human customer service.

AI can be a powerful companion, automating day-to-day tasks and managing virtualised networks while maintaining a culture of human-driven collaboration. Telecoms is already the industry most likely to have reskilled part of its workforce, with 23% having retrained at least one group of employees, compared with 13% in other industries, and Telefónica currently boasts Europe’s largest internal retraining programme.

We’re still in the very early days of AI, and it’ll take some time before suppliers fine-tune their product offerings and commercial models and decide how personal data can be used securely. As everyone waits on the many ongoing legal cases against GenAI providers to be settled, it’ll be some time before these technologies make significant waves in the industry beyond the current hype. Human oversight remains and will remain essential to guiding AI in the right direction.

Nevertheless, there is a remarkable opportunity for vendors to offer AI-powered solutions, trained on organisational data for distinct business use cases; AWS found that only 15% of telcos are looking to build their own models, with the rest opting to feed their proprietary enterprise data into off-the-shelf solutions – probably a wise move considering the rapid evolution of the generalist LLMs which are outperforming industry specific solutions.

Telcos that embrace AI and LLMs across various aspects of their operations should see their customer satisfaction rates and business efficiency greatly improve, predicting and preventing issues, and ensuring the highest level of network security.

Cerillion’s new AI-powered Enterprise Product Catalogue enables CSPs to rapidly build, test and deploy new products and packages using natural language and image recognition. Contact us now to find out more.