Despite Gains, Carriers Still Grappling With Maximizing AI's Value

Carriers are moving to AI-native networks but are still determining what the new technology's capabilities could mean for them, speakers said Thursday during an RCR Wireless webinar. A consensus emerged that AI success requires that different parts of a company collaborate more closely than in the past.

The first thing carrier chief technology officers should understand is that AI-native network architecture and capabilities “look like,” said Leonard Lee, consultant at neXt Curve. “Edge AI is very different from data center AI,” he said. “The role of the telcos, and the business that they’re in, the economics that they deal with, is ... unique.”

Lee noted that his firm wrote a report on “immersive communications” six years ago for the U.K.’s Office of Communications. Using new technology from Apple, you can convert a 2D image into a 3D image, he said. “Imagine doing that for communications,” he said. “That could be a huge frontier for the telco industry,” he said: “It’s actually mind-boggling in terms of what the implications could be just for communications and media.”

Becoming an AI-native telco requires collaboration among all parts of the company, said Michael Raj, Verizon vice president-network enablement. For example, “We are at the point where commercial teams rely heavily on the insights" from network teams, Raj said. Energy efficiency is “a key topic” for carriers and requires cooperation across the business.

New tech vendors and other companies are emerging and carriers must learn from the experience of their peers around the world, Raj said. “It’s very important for us to be up to speed,” he said. “Gone are the days when you can only do experiments,” he said. “We are absolutely tying a lot of these AI-driven initiatives to the value that [they] can bring,” he said. Lots of potential use cases are emerging. “It’s coming from every direction.”

For example, Verizon must assess the strategic importance of a use case and develop a “quick map" of its "value ... for us,” Raj noted. In addition, it's important to assess a use case's technical feasibility. “It can be a great use case with a lot of value, but if it does not have the technology [behind it], we would have to wait.” Carriers also must consider a case's potential risks, he said.

Raj said early use cases for AI include monitoring fiber builds and preventing fiber cuts. AI also has uses in predictive maintenance, with the goal being reducing network downtime. The technology can also help predict energy costs so that there are no surprises, he said. In addition, AI can improve a network's energy efficiency, he said. In customer service, subscribers have rising expectations and they want answers quickly, he said. AI can play a role there.

The challenge is getting "multiple teams working together” using “the same high-quality data,” said Louis Powell, who leads AI initiatives at GSMA.

“The way you start is from the bottom up,” he said. There’s a lot of focus on energy and operating costs and “how to do more without increasing” capital expenditures. Carriers are also using AI to make the work of data centers, field engineers and office staff more efficient, he said.

“The bit that really excites us the most is how do you capture more value through AI -- what does AI allow you to do differently,” Powell said. We’re not seeing that use case “at scale” so far and carriers are still trying to understand ROI, he said. Carriers in some countries are also involved in helping train large-language generative AI models using different languages, Powell said. “Trying to interact with ChatGTP and the other models in Kazakh isn’t a great experience and that’s because the model hasn’t been trained.”