At a well-attended event in London, Rakuten and friends offered a perspective on the reinvention of telecom.
Rakuten Symphony Managing Director Nastasi Karaiskos started proceedings with a roll-call of the innovations that could all be part of telecom’s reinvention, from 5G to Quantum Computing – but right now, AI trumps them all. While most telcos are getting their arms around what it means to be cloud-native, Rakuten has already created an AI-native telco. Not only that, but Rakuten’s use of AI is pervasive, passing way beyond the confines of Rakuten Mobile and deeply interwoven with its wider Rakuten Group activities, in media, finance, travel, insurance and more.
Rakuten is already using AI-powered chatbots to free up human agents to deal with more complex customer service issues. AI is used in ensuring the security of its network, detecting patterns of anomalous or suspicious network activity. And AI in increasingly part of its real time optimization of its network, including the RAN.
Per Rakuten, Reinvented Telecom is AI, Cloud and open and disaggregated networks.
Not so fast…
Former Three UK COO, Graham Baxter, has not seen much evidence – yet – that telco is embracing the fundamental change that this represents. Part of the reason is the daunting challenge of figuring out how to fit radical new ideas and practices and technologies into the context of existing ones – and a network that’s already built out and providing services day in, day out. Ah yes, the L-word: legacy.
Not for the first time we hear that telcos aren’t set up, organizationally, to do the sort of horizontal and holistic/systemic thinking that’s now not only possible but also necessary.
Telecom customers (that’s all of US!) are hard to please. They care more about their devices and applications than they do about the network that they both depend on. And going forward Baxter sees less opportunity for telcos (we’ll assume he means mobile telcos) to differentiate on a network basis. Instead, the three-part recipe for success is a great customer experience, at the lowest cost-per-bit delivered, and some sort of differentiating services on top.
That prompted interesting in-room discussion about the usefulness (or otherwise) of certain metrics (including cost per bit transported) – a topic we’ll return to.
Regardless, a “fundamental shift” is necessary, in order to have ny chance of exploring new potentialities.
Living AI-Native
Madhu Medithe, Chief Data & AI Officer at Rakuten Mobile gave a sense of how life works in an AI-native telco: a process of continuous model training, based on many continuous pipelines of streaming data, pooled in a central data lake, with multiple AI applications feeding decision-making processes from network site planning to offer personalization. He emphasized the need for low level data to be available in order for AI (ML) to work to its full potential.
Where brownfield operators should have an advantage over their greenfield competitors is in the massive amounts of existing data they already have that can be used to build AI models.
Rakuten has plans to use a RAN Intelligent Controller (their own) with NWDAF and AI-based rApps to do non-real-time (SON-like) adjustments to network configuration, though near-real-time uses of xApps remain in early-stage development. There’s some essential digital twinnery in there too, but even Rakuten is not quite ready to let the machines drive the network – having them operate the satnav will do for now.
You Ain’t Seen (Heard/Touched/Smelled) Nothin’ Yet…
Prof Rahim Tafazolli from the University of Surrey, Europe’s largest academic telecom research team provided a view of what’s waiting in the wings once 5G finally takes a bow. It’s satellites, it’s sensory, it’s 6G, it’s 3D, it’s more AI (of course), and much else besides – did I really hear teleportation? Nonetheless, the Surrey unit’s close links with industry should help weed out the theoretical from the real world usable before things get too over-excited.
Buckle Up, Knuckle Down
Prodapt’s Head of Consulting bought things back down to earth by advising the audience that AI-native telco isn’t happening anytime soon for most telcos. A ten-year transition is more likely, with some tough organizational re-programming required. Also that in any sort of 5-level autonomous networks scale, “most telcos are at level 1. Some mobile operators are at level 2.” Ouch. But at least there’s some progress. He cited at least two operators where there’s been precisely the right sort of step change in attitude and ambition (and they weren’t Rakuten and Dish). And the future for AI and telecom isn’t a single all-powerful AI, but tens (hundreds?) of “micro-brains”, all beavering away at their specialized jobs.
The telecom industry seems to have a lot of good answers, but progress depends on more telcos asking the right questions. On the evidence in this discussion, it’s not clear that enough of them are.
Appledore analysts will be at TM Forum’s Digital Transformation World in Copenhagen June 18-21th 2024. If you’d like to continue the discussion there, reach out!