What? Why? How? Who? Four fundamental aspects of telco AI

By James Blackman July 3, 2024

Collected at: https://www.rcrwireless.com/20240703/fundamentals/what-why-how-who-four-fundamental-aspects-of-telco-ai

France-based operator Orange shared hands-on experience of bringing artificial intelligence (AI) into its network operations last month at Telco AI Forum 2024, hosted by RCR Wireless. In interview, Alexis Koalla, director of operations strategy and transformation at the firm, explained the logic (the what and why) to apply AI to network operations, to modernise base-level telecoms infrastructure in time for the AI age, and also the techniques (the how and who) to both remove humans from the management chain, where AI performs better, and reposition and recruit them to extend how AI-enabled networks serve and exploit the top-level explosion in digital services.

RCR Wireless could write around Koalla’s responses in the interview session, but there is really no point; his answers, edited below for language and brevity, explain these points very well, and do not require further editorial input. The full video interview with Koalla is available here; otherwise his point about the cultural mindset and issue of trust is, arguably, the most profound. “[The] challenge for ops teams is to understand… the power of AI, and where it is better, and to really understand [its logic, too], and to [trust its] decision – so it is not just blind trust in AI,” he says. The other opportunities and challenges with so-called AIops in telecoms seem plain, and are well-presented below; all the quotes are from Koalla.  

1 | What? Networks and culture

“Orange is setting up a telco as-a-platform strategy, based on cloud-native [principles]… Cloud-native is the present and the future – because we are moving from a vertical integration… with our vendors to a horizontal one… today with 5G rollout, and because… AI is coming, [and] almost here, [and] we need to consider [its] impact… [This] cloud-native… platform-ization [is] to expose APIs to enable customers to use [the network features]… GitOps is… the foundation of [this] – the… source of truth [for] lifecycle management of [network] assets and… functions. With Kubernetes and CI/CD… we are able to deploy, test, integrate network functions from vendors… to accelerate 5G rollout… What we are doing is new but it’s [also] not-quite new, because we are… applying what we’ve achieved in IT to networks. [It] is not just a technology, [but] a cultural mindset shift.”

2 | Why? Efficiency and innovation 

“The first [objective] is cost reduction, mainly around op-ex. AI will enable significant benefits for that… The second is… to create new value. [Because] otherwise you’re just building a strategy to reduce cost… AI [can] open the business to think about new services, solutions, functionalities, and so on… What we are doing today is ensuring the efficiency of AI on two processes related to the operational part: building new functionalities… [and enabling] fault management… Concrete use cases for AI… are root cause analysis, anomaly detection, auto healing, fault management, service fulfillment, and service assurance; and also, on top, network optimization… [of] energy consumption, bandwidth [usage], and server / equipment [usage].”

3 | How? Humans and machines

“[With] closed-loop [automation] we [want] to reduce or remove human bottlenecks in our processes – [so we can] be faster… [We will] leverage AI… with the right tools and right partners – because we’ll not build everything internally – to [ingest data], extract data [about] an issue, interpret and … filter alarms to analyze and make the right decision… [The] challenge for ops teams is to understand… the power of AI, and where it is better, and to really understand [its logic, too], and to [trust its] decision – so it is not just blind trust in AI. Animal detection, say, is something AI will beat humans at; the same with root cause analysis. Because AI is able, with machine learning, to analyze big data and make decisions faster. AI will not replace people, but it will augment people – so they move faster to get insights and take decisions.”

4 | Who? Training and recruitment

“You can try to upskill your internal experts – and Orange has a lot of them – but it could take time to upskill everybody to reach a certain level with AI. You can also recruit experts from outside. But then you have the culture of the company [to contend with]. So even if you are an expert in AI, it could be a challenge to join a new corporate culture. So at Orange, we are trying to do both – to upskill our current experts… and to upskill [new hires] on the culture of the company. We are starting to see the first outcomes [of that] – to have a kind of continuous learning strategy, to train the right people just enough [and] just in time so… In this way, we’ll be able to keep our experts and make them grow for the company needs for the challenge we are embracing.”