The transformative effect AI is having on the telecoms sector

We consulted industry experts to learn how AI and automation are being used in networks right now, and to explore where they might take the industry as we approach 6G.

Ever since Chat GPT smashed onto the scene the whole world has been a flutter, speculating on the potential impact a new generation of impressively intelligent AI systems might have on the world – both in a good way and in a less good ‘metal skeletons nuking the planet’ way.

The strand of AI that Chat GPT and its conversational bot ilk sits within is called generative AI. This subset of AI has perhaps so sharply grabbed the world’s attention in particular because it so effectively mimics human conversation and even writes long form pieces of content to a remarkably passable degree (though there are famously a litany of glitches and weird results).

Worries around generative AI and what it might do to the job markets, and more philosophical anxieties about what a future artificial general intelligence might do if it were ushered into being by tech firms ploughing their wealth into developing the technology as fast as possible soon became widespread. Consequently, governments are now having a look at what guardrails might need to be put in place as the technology continues to develop.

Regardless, the cat is out of the bag and AI development doesn’t seem likely to slow. It will surely create some winners and losers and the net impact will depend on what area it is applied to. In telecoms there are already automation systems widely deployed, but many see AI taking a much more prominent role in networks as the industry progresses to 6G.

For instance, SkyQuest projects that the artificial intelligence in telecoms will attain a $10.4 billion value by 2030. Such things are impossible to gauge precisely but, taken as a rough estimate, that’s obviously a lot of money.

To get a sense of how AI and automation is being used specifically in the telecoms sector, and how it might be applied in future networks, we assembled a panel of industry experts to pick their brains.

Automation versus AI

When discussing the array of technologies and applications loosely grouped under the AI banner, there can be a tendency to use some of the buzzwords interchangeably. With this in mind, to start with let’s define some of the difference between automation and AI in a telecoms context.

“AI excels in complex applications; automation keeps things running smoothly,” says Kelvin Chaffer, CEO at Lifecycle Software. “AI is the brainiac of the duo, mimicking human intelligence with tasks like decision-making and learning. The amount of data gathered by telecoms is immense. Artificial intelligence can be utilised for more applications to make the best of the data, including predictive analytics, network optimisation, intelligent customer service, and fraud management. Automation, on the other hand, is the workhorse. It keeps the gears turning efficiently and effortlessly, handling repetitive tasks like simple customer support queries, onboarding, network provisioning and maintenance and other routine tasks.”

Niall Norton, Division President, Amdocs Networks expands on the distinction: “Automation is about making something happen (an action), whereas AI is about intelligently deciding what it is that you want to make happen. AI is challenging because the “intelligence” requires historic event correlation (prior experiences and trained outcome models) to be blended with intelligence on causation, which is more subtle and challenging. We do not believe that AI and automation should be pitted against one another but rather utilised in a symbiotic manner.

“Automation is intended to expedite both service delivery and mean-time-to-repair (MTTR) operations. AI and ML can enhance both CSP operational capabilities to increase the rate at which services are delivered to end customers, the number of concurrently operating services, and the average time required to restore a degraded or compromised service. AI/ML enables the minimisation of human intervention to allow domains correlated data insights to be used and leveraged back at the closed-loop framework to make orchestration-appropriate decisions regarding the modification of the service composition.”

AI, automation and the future of network management

Stefano Capperi, Service Assurance & RAN Automation Portfolio Lead, HPE adds: “AI and automation, while often complementary, have distinct applications within the telecoms sector. Automation is geared towards streamlining repetitive, routine tasks, improving efficiency and cost-effectiveness. It is particularly influential in managing network infrastructure. From the core network to the Radio Access Network (RAN), automation can implement end-to-end process optimisation, reducing the need for manual intervention and subsequently enhancing productivity by reducing operational expenditure, a significant boon for a capital-intensive sector like telecoms.

“Meanwhile, AI moves beyond mere task automation to include data-driven inference / prediction, decision-making and predictive analytics. AI is capable of understanding complex data patterns, predicting future scenarios based on this understanding, and making informed decisions. It has considerable applications in the telecoms sector, especially in the realm of anomaly detection within network operations. Using machine learning, AI can identify potential issues and facilitate their proactive resolution, contributing significantly to reducing downtime and improving service quality.

“So, while automation plays a crucial role in executing pre-defined tasks without human intervention, AI enables learning and prediction from intricate data patterns, thereby enhancing the telecom industry’s proactive and adaptive capabilities. This distinction between AI and automation helps telcos strategically apply each technology for maximum benefit.”

Broadly then the distinction seems to be about automation removing the need for human labour in some pre-defined tasks, while AI can attack more complex challenges such as learning from and predicting what’s happening in a network.

What problems can automation solve in the telecoms industry?

With that distinction established, let’s zero in specifically on what pain points automation can help telcos with and where it can provide new opportunities.

“Automation in the telecoms industry addresses the problem of manual and reactive network management,” said Sylvain Nadeau, Director Strategic Innovation and 5G Centre of Excellence at EXFO. “Traditional network management approaches often rely on manual intervention and reactive troubleshooting, which can lead to delays in issue detection and resolution, increased downtime, and suboptimal network performance. With closed-loop automation, these problems are mitigated by creating a proactive and dynamic system. The network is adjusted based on ongoing monitoring and analysis, leading to optimized performance, improved efficiency, and enhanced customer satisfaction.”

Norton adds: “Networks are increasingly complex, and the quantum of services to be launched and updated is growing very rapidly. This is now approaching a point where human beings managing the related systems are becoming the limiting factor in terms of velocity and speed of service management. Put simply, no matter how many humans are involved, business needs are outstripping traditional processes’ fitness for purpose. This is not a telco-specific phenomenon and has emerged in other industries already, for example in high-speed financial markets trading.

“Automation is designed to improve the efficiency of instantiating a given set of solutions/services comprised of resources from multiple CSP technology domains. As wireless/wireline convergence becomes essential, end-to-end services will include resources from all CSP domains: RAN, transport, and all of its sub-domains (optical, IP, Fibre, Copper, Satellite, etc.), 5G Core and related private cloud offerings. Automation seeks to reduce the operational complexity of creating an end-to-end service or solution involving multiple domains and requiring BSS/OSS capabilities incorporating all domains and associated technologies.”

Capperi points to capital intensity as a key aspect of network automation: “Automation offers compelling solutions to some of the most pressing challenges telcos face. Among the most formidable of these challenges is the management of capital intensity. The continuous doubling of network traffic approximately every 18 months necessitates the need for additional infrastructure. This infrastructure expansion incurs significant capital expenditure without any visible signs of decreasing. Automated systems can play a significant role in mitigating these costs by enhancing resource utilisation and reducing manual intervention, thus enabling effective scaling of infrastructure.”

Automation and the secondary device market

Brandon Johnson, SVP Global Engineering and Automation, Assurant points out a use case in the refurbished device market. “One example where automation is being used within the telecoms industry is by increasing the supply and quality of refurbished devices into the secondary market. With supply chain constraints continuing to hamper the availability of new devices, wireless providers, OEMS and retailers are keen to increase their inventories of repurposed devices. AI-driven automation tools are being used during the device repurposing process for data erasure, quality checks and cosmetic evaluation. This not only increases the efficiency of device processing, repurposing more devices in the same amount of time than if done manually, it also ensures it is done to the highest standard—helping to fulfil the demand for quality, pre-owned devices, faster.”

Thierry E. Klein, President of Nokia Bell Labs Solutions Research lists some additional applications of automation in telecoms: “The goal of automation is to put in place enablers that increase agility and efficiency, improve security, and guarantee zero-error, low-cost operations capable of supporting dynamic business demands. Automation can:

– Increase network efficiency and lower operational costs through orchestrated automation across network domains, mainly driven by machine learning.

– Reduce downtime by deploying a ‘zero-touch’ network, where human errors are practically non-existent, allowing companies to deliver an extremely high and consistent standard of service across all levels

– Boost productivity by automating repetitive tasks thus freeing the workforce to focus on other parts of the business that need attention.

– Improve performance through real-time monitoring of networks, deploying network updates, minimizing system outages, and identifying and resolving security issues.

Automation can also help carry out an array of other tasks like managing inventory, collecting network data, ensuring compliance, and updating software.”

How is AI used in the telecoms industry right now?

The current wave of hype around AI has led all sorts of industries to at least start thinking about how it will effect them and how it can be used to their benefit. While a lot of this is in its nascent stage and could be described as somewhat conceptual, it is being adopted by the telecoms industry in practical ways right now. Leaving the future gazing aside, we asked our panel for some real world examples of AI in the telecoms space.

Kelvin Chaffer, CEO at Lifecycle Software told us: “There are plenty of use cases for AI in telecoms, and there is still a lot of space for innovation. We’re crunching data with intelligent network analytics, retaining customers with AI-powered chatbots, and optimising network performance thanks to AI. An example from Lifecycle Software is how we can monitor network events and use AI algorithms to pinpoint crucial moments within the customer journey and deliver preemptive messages suited to the customer’s needs.

“Another key use case is monitoring large volumes of network data and deriving insights for proactive network monitoring, capacity planning, and predictive maintenance. It helps identify patterns, detect anomalies, and predict network failures for better decisions and security.”

Predictive maintenance and customer service

Capperi highlights the predictive capabilities of AI systems plugged into networks: “AI continues to dramatically transform how telcos operate, offering a suite of solutions that go beyond traditional operational capabilities. AI-based predictive maintenance uses machine learning to anticipate potential equipment failures, thereby minimising network downtime and enhancing overall reliability. AI also finds applications in customer service with AI-powered chatbots and virtual assistants that can provide round-the-clock support, thereby reducing the burden on human agents and improving the customer experience.

“In network management, AI-powered tools provide real-time insights and predictive analytics that aid in efficient network traffic management and ensure optimal network performance. In terms of revenue generation, AI can predict customer behaviour, enabling the delivery of personalised offers that can potentially increase average revenue per user (ARPU). These examples underscore the wide-ranging utility of AI, enhancing not just the operational aspects but also customer engagement and revenue optimisation, thus playing a significant role in the telecom industry’s ongoing evolution.”

Norton adds: “Today, we can classify AI/ML as a vitamin supplement that enables specific domains to improve their operational methodology for discovering service assets, implementing customer services, ensuring observability, and optimising the resources used to produce the end-customer service. There are several examples of AI/ML being implemented in this manner.

“First, the RAN domain is progressively transitioning from hardware-attached services to cloud-native service functions. This is to deliver either vRAN/O-RAN enriched with RAN automation capabilities, with Service Management & Orchestration (SMO) functions supported by the RAN Intelligent Controller functions notably for both spectrum usage and customer service optimisation and configuration. Observability and automated service repairs can be ensured with AI/ML capabilities. Similarly, NWDAF is intended to actively monitor the 5G Core components to establish reporting and predictive analytics that would be used to influence the composition of given 5G services and anticipate changes that could be dynamically learned.

“Multi-Domain Service Assurance is intended to capture data from each domain, assimilate data using Machine Learning, construct KPIs and KQIs using AI, and share insights with the Multi-Domain Service Orchestration platform. However, it should be noted that operators approach these initiatives in different ways and at different rates, but they do provide a decent overview of the market situation. We are convinced about the transformational impact that AI is having and will have, but we remain open-minded on the speed of adoption because marrying correlation and causation is challenging and the costs of creating AI models have yet to mature.”

So that’s a snapshot of some ways AI is currently being used by the telecoms industry today – but what impact might it have in the years and decades to come?

AI, 6G, and beyond

How AI is used will of course change as technologies improve, new use cases emerge and business reorganise themselves around a changing technological landscape. For the telecoms industry, a key focus point will be how 6G might develop in relation to evolving AI systems.

Nadeau says: AI is expected to play a crucial role in transforming network operations and capabilities. 6G will be a cognitive network that self-adapts using machine learning (ML). There will be a ML plane alongside the existing user and control planes. AI algorithms will analyze diverse factors such as user demand, traffic patterns, and geographic conditions to optimize the placement of network elements, antenna configurations, and coverage strategies, resulting in more efficient and cost-effective network infrastructure.

Klein adds: In 6G, AI will go beyond enhancing the network to becoming a foundational technology directing how the network is designed, built, operated and optimized. 6G will take a clean-slate approach that will do away with complexity and let AI aid in determining the best way to communicate between two endpoints in an energy-efficient manner.”

AI at the edge and the promise of dynamic network slicing

Chaffer believes having AI operating at the edge will be a key component of 6G: “As we gear up for the 6G era, AI takes centre stage. A potential application is AI at the edge for lightning-fast decision-making. AI will enable dynamic and intelligent allocation of network resources, from bandwidth allocation to energy efficiency. 6G will maximise the ability to segment the network’s service into logical slices. Each slice supports a specific set of business cases and thus offers a varying quality of service (QoS) depending on the demands placed on it by the business case. This slicing ability, coupled with AI speeds, sets the stage for innovative use cases.”

Capperi says AI will also have a role in sustainability in the 6G age: “Another area in which AI will play a significant role is sustainability. Most telecommunication service providers have made commitments to dramatically reduce their carbon footprint over the next decade. 6G has specific working groups driving sustainable operating models for networks. AI will be used to reduce the energy consumption in both the RAN and Core. Telecommunication networks will increasingly be driven by control systems embedding AI.

“The relationship between AI, automation, and telecoms is multifaceted and rife with potential. As the technological landscape continues to evolve rapidly, one thing remains certain – the continued convergence of these advanced technologies and telecom will drive the future of innovation and progress.”

The role of AI/ML in shaping next generation communication networks

Finally, Norton adds: “6G is likely to increase the complexity of wireless networks and support a new level of capability, most notably the provision of THz communications using quantum computation. Consequently, big data analytics will necessitate a more pervasive AI/ML. AI and ML will enable the microseconds observability required to achieve the objectives of 1Tbps peak rate, 1ms end-to-end latency, and 20-year battery life for the next iteration of communication networks.

“To achieve faster rates and lower latency performance gains, an extremely efficient network will be required, with AI/ML supporting the fully automated OSS layer to dynamically allocate resources, modify traffic and priority flows, and process signals in an interference-rich environment. Consequently, we can anticipate the consideration of genuinely novel concepts of dynamic security, network segmentation, and private network management.”

Depending on who you ask, AI can be described as a bright new opportunity that will help propel businesses and societies to great new things, a big problem on the horizon we are all going to have to work out how to live with, and anything in between. In the context of telecoms networks there are clear cases where automation and AI is being used already to streamline various operations of 5G and 4G networks. Furthermore, the consensus of our panel is that AI will be heavily integrated into 6G when it arrives, so let’s hope we have sturdy guardrails in place by then.

Get the latest news straight to your inbox. Register for the Telecoms.com newsletter here.