Orange Business leverages network smarts to monetize AI

Telcos seeking a profitable position in the AI value chain are at a critical juncture, according to Usman Javaid, Chief Products and Marketing Officer, Orange Business.

“I think it's a moment of winners and losers. Those who see AI as an opportunity to grow again are the ones who are going to win, and those who miss the boat on AI are going to completely lose,” he says in an interview with Inform.

Javaid points out that GenAI companies are shifting their focus away from the delivery of commoditized language models and onto the delivery of services.

“The value chain is massively shifting, and it's shifting upwards. If you listen to Open AI and Anthropic … they say they are a product company. They are not a model company. They are trying to fix specific problems in healthcare, in transportation and so on,” he says.

Communications service providers (CSPs) therefore need to urgently consider how they can make money out of the coming wave of AI-driven service innovation, argues Javaid. Up until now the focus of much of the industry has been on using GenAI to drive internal operational efficiencies.

“I'm not worried about us being able to use AI in our operations, in our networks. I'm pretty sure we will do a cracking job at that. For me, it's about not repeating history,” says Javaid. By this he means telecoms infrastructure enabling service growth without taking a share of the spoils.

“If you look at the mobility wave, we created so much value for the industry,” he says. “All the Ubers, the fintechs, everything is riding that wave that we created, but we were not able to capture the value.”

The question for CSPs is where they fit in a fast-evolving AI value chain.

Last year Javaid co-authored an article that argued the case for building networks for AI applications, which stated that “scaling GenAI deployment requires network and compute reconfigurations for balancing centralized training and rapid edge AI inference. How the network is built to support AI-enabled applications needs revisiting urgently.”

Compute on the fly

Developments over the last year have strengthened his conviction that telecoms networks designed for AI through to the edge can enable revenue growth for communications service providers (CSPs).

In particular, he points to improvements in the efficiency of GenAI language models, which now require fewer resources to operate, and which shift the cost of AI away from training models and towards running them.

"Compute needs are moving more towards inference because the models are becoming more reasoning based and need to think on the fly. So, it's compute on the fly,” he says.

“This is very close to the service provider space because running [the models] requires networks. It requires real time performance. The more it becomes immersive, the more the network will have to be able to offer the right level of latency, privacy, compliance, and so on.”

Orange has opted not to build language models. Instead, it favors partnerships.

“Companies that are … GenAI native … know their stuff and I think we should leverage that,” says Javaid.

Earlier this year, for example, Orange announced a partnership with France’s AI powerhouse, Mistral. Together they will “assess the impact of large-scale, massive use of AI on telecommunications networks worldwide” and “define technological roadmaps to build the networks of tomorrow and address challenges related to connectivity and GPU availability,” according to the release. It has also teamed with Open AI and Meta to expand opensource AI models to African regional languages.

Monetizing AI

Orange’s aim is to give customers “smooth and efficient access to advanced AI solutions,” says Javaid.

In addition to gearing its networks to support AI, it is pitching GenAI services to enterprises through Orange Business’ Live Intelligence platform.

Live Intelligence started out as an internal platform to avoid the mushrooming use of shadow GenAI applications. It proved to be such a popular problem-solving and optimization tool with 60,000 of Orange’s employees that it became a product.

Now Orange Business sees it as the means to capture a share of the burgeoning AI value chain.

“As we shift into a platform approach with Live intelligence we capture value through the ecosystem, so that once the industry moves ahead and enterprises build new value propositions, we have the fair share of value,” Javaid explains.

Building GenAI solutions for enterprises “requires a lot of hand holding, and it requires a consulting led approach; helping customers to understand the use cases,” he adds. He cites the example of a GenAI-based tool Orange Business built for a French energy company, which enables field technicians to automatically generate reports upon completing a task, while also respecting the company’s data sovereignty requirements.

Some “customers … are very sensitive to where their data sits and where are the models, and who builds those models,” explains Javaid.

As agentic AI develops and customers seek outcome-based AI services, Javaid sees Orange bringing connectivity and security to future partnerships.

“We connect enterprises around the world. So, who knows better than us their specific … [networking] needs?” says Javaid. “We are one of the serious players in cyber security. We know the potential cyber threat which may potentially emerge … [and where] based on the data that we have.

“We will be able to build agents in those spaces, and then, of course, partner with others who know things around their specific domains,” he explains. “We have got a lot of understanding of mobility and but then Microsoft [for example] have got understanding of how people collaborate and putting the pieces together, you would be able to create something … quite powerful.”