They can choose routes that make them essential drivers of the AI future
In sum, what to know:
Current success can lead to future success –Telecom has succeeded in its substantial, practical AI implementations, but so much more is possible through strategic investments and evolving partnerships.
Cloud native for flexibility, scalability –The move to cloud native will make it possible to handle the dynamic workloads and the real time processing that come with AI.
Compute, fiber optics, and edge – Investments in computing resources, edge AI infrastructure, fiber optics, and virtualized network functions lay the foundation for AI as a growth avenue.
In many sectors, AI projects die in pilot purgatory, but telecom has managed to make substantial, practical AI implementations, particularly in network optimization, predictive maintenance and customer service. AI-powered chatbots and virtual assistants have improved customer satisfaction and reduced costs; AI-driven predictive analytics have helped automate resource management, network performance and operations; and success has been realized in fraud detection, network security, and revenue assurance.
But Cédric Foray, EY Global and EMEIA Telecommunications Sector Lead, told RCR in a recent video interview that there’s so much more telcos could be doing. “They are currently at the intersection of AI and infrastructure decision-making,” with some looking to adapt their networks to investment roadmaps, looking at the full array of AI services they can utilize internally, as well as extend to their customers, such as connected devices and IoT-related connectivity services, GPU-as-a-service, and more.
Looking to the AI future, telcos are certainly making investments in foundational areas. They’ve virtualized and cloudified, and now they are moving to cloud native functions and apps that will enable them to deliver more value through their AI investments. “Cloud native greatly simplifies the management of the data center, making it possible to handle the dynamic workloads and the real time processing that come with AI,” acknowledged Foray. “Cloud native provides the scalability and flexibility that will open the door to AI-driven improvements in customer care, network and back-office use cases.”
In addition, telcos are investing in computing resources (XPUs and GPUs for complex AI models), edge AI infrastructure (for low-latency AI processing close to the user), and upgraded network infrastructure to handle AI workloads (fiber optics, and virtualized network functions). Increasingly, leading telcos like Verizon, AT&T, Vodafone, Deutsche Telekom, Telefónica, KDDI, and others are investing in partnerships with cloud companies and hyperscalers like AWS, Microsoft Azure, and Google Cloud.
Story continues below video
These companies are tapping hyperscalers’ computational power, AI expertise and analytics, while providing to them the fiber and wireless infrastructure that transports data among data centers. Telcos also provide the dark fiber and network services that bolster redundancy and the ability for cloud and hyperscalers to scale capacity. This is all in addition to the cell towers and fiber optics telcos provide, as well as the regional expertise and last-mile partnerships that are essential in expanding AI infrastructure and services.
Edge computing is the next frontier, according to Foray, especially as AI applications that require low latency, like AI inferencing, autonomous and the IoT, continue to evolve. “What telcos need is a more strategic approach…to clearly define business objectives, establish clear metrics for success, and do everything they can to assure cross-functional collaboration and end-to-end views,” he said.
He sees four key obstacles that must be overcome if proof-of-concept pilots are to be converted into AI successes:
- Ongoing reliance on legacy network infrastructure, “which is not designed for the real-time processing, or the massive data requirements and integration that modern AI applications demand,” explained Foray.
- Data silos and quality of data. “Telcos have vast amounts of data about customers, network performance, operational networks — all trapped in silos because of fragmented systems with incompatible formats that hinder the training of AI models. This is where the investment in integration is going to have to be,” said Foray, noting that modern, cloud-native AI applications are API-driven and modular in design.
- Talent acquisition and retention— the persistence of legacy systems and infrastructure means specialized talent is needed to maintain those outdated, but still critical systems. “What’s needed is talent that understands not only AI, but the unique characteristics and challenges of telecom. Upskilling programs and initiatives to get a pipeline of talented people are an absolute must,” said Foray, noting that major tech companies and hyperscalers are tapping the same limited pool of talent (often with much more attractive salaries and benefits).
- Too many point solutions for emerging technologies, which further fragment data and further complicate simplification and integration efforts.
As these obstacles are overcome, Foray believes the industry will move more successfully toward generative AI, AI-native networks, AI-RAN, and monetization via GPU-as-a-service. With most telcos actively deploying AI projects for operational efficiency and CX boosts, they will amass vast amounts of network and user data, which will be a strong foundation for future AI applications and analytics.