Orange bets on AI for network optimization, automation

Pan-African telecom operator Orange has been using artificial intelligence (AI) to optimize its network in a number of African countries and is investing in AI capabilities to increase automation and improve customer service.

Brelotte Ba, deputy CEO of Orange Middle East and Africa, spoke to Connecting Africa on the sidelines of Africa Tech Festival in Cape Town, South Africa. He said the operator is using AI algorithms to balance network capacity and coverage, predict equipment failures and automate network maintenance to ensure zero downtime.

"We have a challenge to build the network and to [increase] capacity and coverage, because usage is increasing. We're constantly adjusting the capacity to make sure that we're giving the customer the right service, the right bandwidth and also good coverage," he explained.

"We have our algorithm using machine learning to see – based on the demand, customer needs and the traffic – where to invest to make sure that we have a good balance between the coverage, the capacity, and [the ability] to provide the best quality of service," he added.

Orange calls the AI solution "Smart Capex." It was first piloted in Spain and has now been rolled out in Côte d'Ivoire, Senegal and Mali in Africa, as well as in Jordan in the Middle East, to help optimize network capacity and streamline the investment process.

Related:Orange, Open AI and Meta partner to train AI models in African languages

"The beauty of AI is that you have many real-time use cases combining all this data, instead of having people looking at each and every piece. That's the biggest impact of AI," Ba said.

"We're using it to constantly adjust real-time network capacity and the demand, and also anticipating on where to put the network," he said.

Orange operates in 18 countries in the Middle East and Africa (MEA) and has 156 million customers in the region.

He clarified that when it comes to the "Smart Capex" solution, the aim is not only to save costs but also to invest in the right places.

He said the benefit is direct in terms of quality of service, and in terms of managing energy consumption.

"You're powering equipment and [users] are consuming [services], but if you can predict what the traffic will be, then you can adjust the power and say, at this moment of the day I don't need 100% power, maybe 40% will be okay, and then you adjust it," he explained.  

"All the machine learning algorithms are really here to make sure that we're having the right pace in terms of customer needs compared to equipment, and not only on the traffic side, but also on the energy side. So, these are the capabilities that machine learning algorithms are bringing the network," Ba added.

Related:Orange eyes more African 5G launches in 2024

Building a fully automated network

Ba said the second use case for AI in the telecom market is automation.

"Look at the car industry. Many car manufacturers are trying to have the driverless cars, especially in the US, and all this is based on AI and bringing AI capabilities to automate the car. It's the same for [mobile] networks. You have, for instance, use cases on predictable maintenance, using machine learning to see when we may have a failure in one piece of equipment, and then remove it, replace it before the failure occurs. So, in this regard, you will have no downtime," he explained.

"It's, of course, better than just waiting until you have this crash, replace the equipment and you have a downtime and customers are complaining about that. So, this is the idea of zero-touch network, meaning that network can work smoothly, provided that you have the algorithm to give you a sense of where you need to replace equipment," he continued.

Orange is using AI algorithms to balance network capacity and coverage, predict equipment failures and automate network maintenance. (Source: wirestock on Freepik)

"We have had a solid journey because, back in 2015, the idea was just to improve efficiency and have the local competencies when monitoring the core network, and now we can use all the capabilities of AI in this organization to go one step further and automate the whole process of monitoring the core network," Ba said.  

He said that using AI solutions does not remove humans from the equation completely, because a technician still has to replace the faulty equipment, but if an algorithm can detect a fault before the equipment completely fails it will mitigate network interruptions.  

"The idea is to detect it, then to change it and then your network will be self-healing to some extent, because you will not have the downtime. So, there will still be humans, of course, who are here to interact, to change hardware. This should be done, but at least they will not be here just to wait until it's faulty and then change it," he added.

"Automation will bring efficiency and cost savings, etc., but actually the use cases are varied and cover a large panel around empowering people, having better interactions with customers, investing right and driving efficiency," said Ba.

GenAI solutions for customers and staff

Orange is also leveraging generative AI (GenAI) to improve customer service through a knowledge database chatbot called TutoGenius, which was built in partnership with data and AI transformation consulting company Artefact.

Ba said that TutoGenius is a GenAI-powered solution for customer queries.

Although the operator was already using chatbots, the new solution is an upgrade using generative AI so that customers can interact with it in natural language instead of having to have the right prompt to get the correct answer.  

"The beauty is that you get an answer, talking to it as you would talk to anybody, and even if you're giving additional insights, then it has the memory of interacting with you. So that's what we're bringing to the customer relationship," he said.

Orange's AI services are built in via partnerships with companies like Artefact and also through internal teams based in Côte d'Ivoire, France and Jordan.

Last month, Orange also announced a partnership with Meta and Open AI to train AI models in African languages.

Orange has developed a generative AI-powered chatbot called TutoGenius for customer queries. (Source: Freepik)

In addition Orange has launched a GenAI-powered solution for its employees called Dinootoo, which can assist with tasks, generate images and analyze documents

"It has been customized to cope with some constraints that we have in terms of intellectual property or data privacy so that data is not flowing to the Internet. We're using all the LLMs [large language models] that are out there, like Mistral, ChatGPT, Gemini, etc.," he added.

AI adoption and skills development

Ba believes that AI adoption is growing in Africa and across the globe, especially when you consider the millions of individuals and companies already using LLMs.

He referenced The AI Index Report from Stanford University, which found that in 2023, the industry produced 51 notable machine learning models, while academia contributed 15. There were also 21 notable models resulting from industry-academia collaborations.

"It means that companies are driving this [adoption], and we see it very much in a company like Orange where we have all these use cases that we are pushing, that we're investing in, so adoption is already here," he said.

He added that there is a discrepancy between markets, with 61 notable AI algorithms coming from the US, 21 from Europe and 15 from China, according to the same AI Stanford Index.

Ba said there is a big opportunity for Africa to contribute to AI innovations. But for that to happen, there needs to be more training opportunities for young people locally.

"That's what we are doing in the Orange Digital Centers, training them at scale, having the right partners, so that people can know how to make algorithms, to do machine learning, to be trained to be software engineers, software experts," he said. "That's the chance that we have in Africa, with a young population. If we train them, they can come up with very good ideas."

He did, however, note that there is a challenge when it comes access to data.

"There is no AI without data and training. Even if you know the sophisticated algorithm on paper, you need to train it using data. So how do we access the data? What are the sets of data that are available in Africa? Most of the data are in the big platforms. So, we need to also make sure that we can have data that is on the Internet with the local languages so that people can access it. [We need] maybe a global regulation, including Africa, which will give us this source so that people can use it," Ba said.