OXON HILL, Md.—In panels and conversations at this week’s International Telecoms Week, some of the reality of practical AI applications for network operators emerged.
“We are already getting a lot of value from AI,” said Jean-Louis Le Roux, Orange’s EVP of international network infrastructure and services, during a session on network operators participating in the AI value chain.
McKinsey earlier this year had concluded that with the support of AI, network operators are “quickly reaching the point where they can be more precise about their capital allocations to efficiently expand and improve the network.”
Le Roux and other network operator panelists went on to lay out the areas in which they are actually already realizing value from artificial intelligence.
Network operations. “We are already using AI in order to automate the capacity planning of the network, optimizing the network and ensuring the best resiliency,” Le Roux said, adding the Orange also uses AI to monitor the network and manage tickets that come from customers.
“All this, I would say, is improving the customer experience and the quality of service, and optimizing our cost,” he added.
Jeff McHardy, head of international network capacity planning and commercial engineering at Telstra International, said: “The dream is to have a totally autonomous network, right? That’s where we’re driving to at Telstra.”
McHardy said that the company is putting fundamentals in place now to work toward that goal, by cloudifying its network and separating digital infrastructure hardware components and software, in order to capitalize on AI capabilities for automation. He said that the process enables better tailoring of services to customers, monitoring traffic and traffic segmentation, and more effective capacity planning and analytics.
This ultimately means, McHardy went on, that network operators can run their networks more “aggressively”, because they have deeper understanding of their traffic trends and needs. “You get a greater return on your network—you don’t have to hold greater headroom in your network to manage bursts,” he explained.
Security. Amid rising security threats and attacks, Orange is using AI as part of network security to protect its network and its customers, Le Roux said.
According to Verizon’s Data Breach Investigations Report for 2025, there’s been an “alarming surge” in third-party cybersecurity incidents and attacks — a growing threat that’s reshaping the risk landscape for organizations of all sizes. Although generative AI (gen AI) is not yet dominant in attacks, its fingerprints are beginning to show, the report found.
Using AI in this context is, in part, about fighting fire with fire. “The people who are attacking us, they use AI. So we have to react with AI,” Le Roux said.
Understanding customer needs and behavior. Le Roux specifically pointed to the example of roaming as one way that Orange is using AI to better understand customer needs and offer services accordingly. Having a better understanding of customer roaming behavior enables the company to adapt its services and gain additional revenue, he said.
Fighting fraud. Le Roux also categorized one strategy for using AI as protecting existing revenues—specifically, to combat fraud related to voice services. Orange faces multiple types of voice-related fraud: International revenue share fraud (IRSF), or exploiting international premium-rate numbers; caller line identification (CLI) spoofing, so that callers think a call is originating from a legitimate person or business; and origin-based rating (OBR) fraud.
“Here, AI is very powerful, because we can easily detect the fraud and we can stop the fraud,” Le Roux said.
In the future, network provisioning is also expected to be supported via AI. This is a future evolution, Le Roux said, but it is one that Orange is exploring now through proofs-of-concept. He can see a day when “AI may be used to revolutionize the way we provision the network.” Relying on generative AI, he explained, complex configuration decisions and actions could be made by essentially having a conversation with the network: asking about activity and traffic in order to optimize provisioning. “This is very, very promising,” Le Roux said.
