Originally published on techtarget.com

AI and machine learning are having a significant impact on wireless networking and telecommunications for network operators. Various innovations have enhanced coverage, optimized operations and improved customer experience.

At the same time, telecom operator lethargy and pushback contributed to the rise of alternative approaches, driven by the simplicity of alternatives like TCP/IP and Wi-Fi. Other innovations in wireless telecommunications show promise for extending the reach of enterprise operations across a wide variety of use cases.

When 5G first came out, it promised better and faster networks across a multitude of applications. Yet users were frustrated by dead zones that required failing over to legacy network protocols to fill the gaps. The core 5G protocols also enable 5G to be a promising alternative to Wi-Fi in enterprises and can help reduce latency as well as improve resilience. Nevertheless, commissioning a 5G link is still a more challenging proposition than onboarding a new Wi-Fi device.

Recent innovations in AI, particularly generative AI (GenAI), are helping to mitigate these challenges. And just around the corner, next-generation 6G standards that use AI more directly in the network to enhance resilience could help overcome these limitations, thereby boosting enterprise adoption of these cellular telecommunications standards as a viable alternative to Wi-Fi and other internet options.

“Currently, the telecom ecosystem is benefiting from what AI can bring to the table in several ways. Directly related to the telecom systems, operators are now building LLMs [large language models] that can analyze incoming packets to predict maintenance and downtime before it occurs,” said Scot Glover, global leader of technology, media and telecommunications at Protiviti, a digital transformation consultancy.

Another concern is the significant power needed for telecommunications networks. While AI consumes power to perform its tasks, the ability of AI to build and design new radio access networks and manage them more efficiently during both peak and slow times should bring about significant cost reductions over time, according to Glover. These new benefits will lead to better performance, more eco-friendly networks and improved satisfaction for customers of mobile virtual network operators — i.e., service providers that lease mobile infrastructure.

AI is also giving telecom providers an opportunity to deliver value beyond what’s possible with traditional TCP/IP and Wi-Fi approaches. “AI innovations are creating material opportunities for telcos to improve their core business by driving efficiency through autonomous agents running processes, boosting productivity with AI assistants that complement human work in sales, service and the field force, and enabling hyper-personalization through generative AI that produces creative content and powers next-best-action algorithms,” said Davide Bellini, Accenture’s global data and AI lead for the communications media and technology industry.

In the future, he said, this could lead to new business models where telecom service providers combine connectivity with computational power to support enterprise-managed AI that is distributed through edge AI products.

How AI Can Optimize and Improve 5G

AI has the potential to benefit 5G in the following main areas.

Mobile Device Management

The top use case for enterprises lies in improving mobile device management (MDM), traditionally a burdensome process that spans service, business logic, and various mobile devices and authentication mechanisms.

In an era when mobile devices are increasingly used to access enterprise services, MDM is a growing concern. Telecommunications-generated credentials, including subscriber identity module (SIM) cards and, now, electronic variants called eSIMs are increasingly used to support access to enterprise services via two-factor authentication. Yet the traditional SIM architecture championed by legacy telecom companies never addressed the numerous ways these authentication mechanisms could be compromised, including through stolen phones, corrupt agents or poor workflows. AI can enhance visibility into these issues, thereby improving the resilience of enterprise systems.

According to Krishna Iyer, vice president and group CTO at GlobalLogic, a software development firm, telecom networks are undergoing a significant shift from reactive, manual operations to AI-driven, proactive, increasingly autonomous paradigms. Here, generative and other types of AI can improve customer care, network operations, business support and software delivery.

Streamlining 5G Adoption

Another promising proposition is that AI could streamline the implementation of private 5G networks that span private and public 5G infrastructure. These networks face less interference compared with Wi-Fi but have more complex provisioning of new devices. AI could help overcome these bottlenecks.

“Private 5G used to intimidate CIOs because it felt like running your own miniature carrier network,” said Sriram Panchapakesan, CEO of telecommunications, media, technology, energy and utilities at Sutherland Global, a digital transformation company. “AI changes that. It’s the difference between manually flying a plane and having an autopilot, since the network tunes itself, maintains tight security and adapts to your workflows.”

With AI helping to streamline the process, companies can spin up a private network in days rather than weeks. Panchapakesan has seen factories use the technology to connect thousands of sensors, while logistics hubs use it to track vehicles in real time, proving private 5G can be practical and scalable.

Weighing in on 5G management challenges, Iyer said AI can help simplify and enhance deployments through predictive traffic optimization, automated slice management and proactive anomaly detection. Moreover, recent innovations in GenAI can make operations more accessible in ways that enable network and IT teams to troubleshoot and respond to issues in natural language rather than relying on deep expertise.

For example, digital twins and service-level agreement (SLA) recommendation engines can help enterprises right-size network resources to their needs, whether they involve dynamically adjusting bandwidth or providing ultra-low latency for autonomously guided vehicles and equipment. Innovations in agentic AI orchestration could add the ability to diagnose and resolve issues that previously required technical experts or lengthy escalations. “The result is a private 5G environment that delivers higher performance, less downtime and stronger SLA compliance while easing the resource burden on operators and simplifying operations and management for enterprises alike,” Iyer said.

There is also an opportunity to improve the speed and range of 5G infrastructure in challenging environments. “Long gone are the days of sending a text at the football game that never goes through because 90,000 fans in the stadium have completely congested the one tower in the area,” Glover said.

In another example of streamlined private 5G adoption, airports are starting to use 5G rather than Wi-Fi to improve maintenance, luggage tracking and boarding.

Better Telecom Plumbing

The most challenging aspect of obtaining the benefits of 5G is improving the underlying infrastructure. “Telecom data is messy and lives in silos, and integrating AI into legacy operations takes patience,” Panchapakesan said. “It would be ideal to start with perfectly clean data, but that often isn’t realistic. The good news is that AI itself can help clean and prepare the data for other models and agents, and it can enforce governance to keep it clean once you get there.”

But using AI for this purpose also requires earning trust. Network engineers might feel uneasy about AI making decisions that could affect millions of users. Panchapakesan recommends that enterprises allow AI to assist with data preparation and governance, while also designing systems to explain the organization’s choices.

A key value proposition for telecom service providers lies in explaining how emerging 5G infrastructure can enhance enterprise sovereignty and control.

Bellini sees this as a huge opportunity for service providers. “They can combine network transformation with computational infrastructure and sovereign AI, which puts them in a unique position to deliver a truly complete connected computing infrastructure, which drives efficiency and allows for more effective management,” he said.

Enabling Decentralized AI

In the long run, AI also holds promise for supporting telecom companies in becoming providers of intelligence by offering networks optimized for AI traffic to alleviate concerns about AI computing capacity. “Generative AI, in particular, is heavily compute-bound,” said Frank Long, an associate director at Cambridge Consultants. “Running the largest reasoning models efficiently requires water-cooled data center capacity, using top-end GPUs, both of which are currently in short supply globally.”

Telecommunication providers operate numerous data centers that can support mobile edge computing (MEC) locations, which could be repurposed to serve AI systems. However, it would be challenging and expensive to convert those MEC locations to high-capacity data centers because there might not be the local energy or environmental resources to support them, Long said.

He sees an emerging challenge for telecom providers in helping enterprise customers create cost-effective, efficient AI applications, given the inevitable hierarchy of computing resources needed for high-powered, water-cooled data centers, mid-power distributed GPU locations and low-power AI running on devices. On the other hand, the same hierarchy could enable enterprises to use the appropriate amount of computing resources at a lower price overall and with the right latency characteristics.

The situation is comparable to that of a national utility running an AI-aware monitoring and control system that employs reasoning models for situational awareness, processes large data sets collected from the field and continually refines the live operational digital twin of the network. Such a system could be supplemented by local monitoring, aggregation and control, with AI systems running at edge locations maintaining the health of systems in their geographic area, acting as the body to the monitoring and control system’s brain. These could be supported by intelligent sensors using low-powered AI models that work collaboratively to achieve the desired service levels.

This vision presents inevitable challenges:

• How do telecom provider customers ensure that the training data for such a diverse range of AI systems and models is coherent and doesn’t require upgrades that result in highly unpredictable outcomes?
• In a future where AI agents take action, how do you scope security permissions to both empower and control AI systems, especially in highly dynamic systems?
• How can enterprises empower development teams to create optimal solutions that work across such disparate hardware without being overwhelmed by complexity?

“There will be winners and losers in this emerging landscape, Long said. “The winners will shift from being providers of connectivity to being national, regional and global providers of intelligence. It’ll require partnering with AI orchestration platforms and the hyperscalers, such as AWS, GCP and Azure, as both enablers and competition.”

Not a Smooth Road

Numerous hiccups could slow down the adoption of AI for telecommunications across the enterprise. Bellini observed that many telecom companies struggle with limited capital, immature data estates and outdated legacy IT systems. Added to that are the complexities of building the right capabilities to ensure transparency in both AI costs and decision-making.

“Addressing these challenges requires a dual focus, whereby telcos need to modernize core systems and data foundations while also creating governance frameworks that ensure responsible, transparent and compliant AI adoption,” Bellini said.

Iyer also highlighted challenges with fragmented data, AI model drift and the need to support explainability, compliance, accuracy and scalability across mission-critical networks. Across these various contexts, the old notion of “garbage in, garbage out” in IT systems also applies to many types of telecom ecosystems. However, it’s especially relevant for AI agents, which can struggle with poor-quality or incorrect data. In these cases, he said, the transformers used in LLMs could help translate data and improve data quality in more targeted ways than general-purpose LLMs.

“The way forward is to unify the data layer, embed responsible AI practices at every stage and focus on closed-loop use cases like SLA optimization, interference detection and churn prevention, where results are measurable,” Iyer said.

Glover named cost as the most significant issue with AI adoption in telecommunications for enterprise use. “The skill set and technical resources to effectively leverage AI are at a staggeringly high level,” he said. In the meantime, organizations will have to acquire expertise that is costly and in demand.

The Future of AI in Telecommunications

Increasingly, new GenAI techniques are showing promise in helping to bridge gaps across enterprise processes.

Explainability, compliance and sovereign AI principles are also essential to ensure that sensitive telecom data and operations remain secure, transparent and ethically managed.

Emerging 6G infrastructure promises to weave AI directly into air interfaces, protocols and services. This will support new capabilities using efficient semantic communications, the fusion of sensor data, and streamlined integration with digital twins — all building on existing 5G AI foundations.

Debashis Basak, co-founder and CTO at Highway 9 Networks, a mobile cloud provider, predicts AI will support rapid advances in agent-based architecture, self-improving systems via reinforcement learning, and domain-specific foundation models. These agentic approaches will not only detect and fix issues autonomously but also optimize radio-frequency planning, slice management and service assurance in real time. At the enterprise edge, on-premises AI inference could unlock use cases such as real-time video analytics, robotics and manufacturing.

Accenture’s Bellini believes agentic AI will continue to expand rapidly across telecommunication use cases. “We’re already seeing telcos push toward transforming their connectivity into fully autonomous networks, and the near future will be less about experimentation and more about scaling these technologies successfully,” he said.

Panchapakesan predicts the next wave of telecommunication will be about orchestration and autonomy. “We’re getting close to networks that can predict where demand will spike and pre-position capacity or spot a security anomaly and neutralize it in seconds,” he said.

Telecom operators and enterprise network managers will test new network configurations in virtual environments before deploying them in production. AI will become almost invisible at the edge and embedded in chips and network nodes, allowing enterprises to run low-latency, high-security apps without worrying about the underlying infrastructure.

Glover expects that 5G grids could become fully integrated into cities and roadways, enabling seamless interaction among humans, machines and nature. Smart cities could manage their power grids effectively and help road and highway agencies prepare for disruptive events. In addition, a common language that enables AI systems to communicate with one another could help agentic AI operate at an extremely mature level, he predicts.

About The Author: George Lawton is a journalist based in London. Over the last 30 years he has written over 3,000 stories for publications about computers, communications, knowledge management, business, health and other areas that interest him.