The big announcement last week from IBM and the GSMA about teaching telcos in the ways of the force – as in generative artificial intelligence (AI) – is not just about telcos. Or rather, it is not just about showing telcos how to use AI to make efficiencies in their networks, and replace call centre staff with chatbots. In the end, as ever, it is about enterprises, and how telcos keep pace with the fierce industrial adoption of an era-defining technology, and how they might actually use it to make their own connectivity services meaningful to new revenue-driving customer segments.
Which is what we want to know, and what we want to talk about. Half way through a discussion with IBM about the initiative, trailed in the telecoms press like an MWC curtain-raiser, RCR Wireless puts its cards on the table: the thing is (for this old hack) that generative AI gets talked about in telecoms in terms of customer care and call centres, mostly, and how to make radio engineering easy for a generation of new hires – and none of that is very interesting. Because the exciting bit is how mobile operators, plus all kinds of others, might serve the rest of the global economy.
So how much is this about telcos getting their AI ducks in order to raise ROI from inward-looking operational efficiencies, and how much is about outward-looking value creation? It might say something, and also explain the line of questioning, that Stephen Rose, general manager for global industries at IBM, is handling the American tech firm’s end of the bargain. He responds: “You know, one of the Indonesian operators asked the other week whether, with a pound to spend, it should go on saving money or making money. And my answer is to spend it on making money.”
He goes on: “It starts with the second goal, right? The telecoms industry is full of optimisation programmes; they have been doing them for decades, and they’re pretty good at it. And we are part of that discussion, of course. But this is more about how their customers expect to receive their service, and how AI makes a difference. If you’re going to use AI for visual inspection, say, then how is that applied to multiple different industries? At some point that seeps back into the way the service is delivered, but it is a growth-enabling initiative more than it is an efficiency-saving one.”
Just to backtrack. The GSMA and IBM have launched a “training programme and industry challenge” to drive generative AI education, experimentation, and innovation in the telecoms industry. The scheme divides into three parts. The first phase, already taking its first cohort, establishes a kind of AI school at regional IBM offices for c-suite execs, mostly at tier-two/three operators, “to understand the techno-economic potential of AI”. It will also extend to developers in different telco departments – “to turn users into content creators”.
The creator-mode is made real in the second phase, which, as scheduled, presents a live ‘sandbox’ experimentation “platform” with predefined use cases in different domains (“customer care, customer experience, and so on”) so developers can get their hands dirty with generative AI. A third phase, in early planning, would build a federated data hub, “almost like a condo”, between non-competitive telco graduates in a region, to share data to build AI foundation models in a collaborative manner that raises the AI mastery of a bunch of tier-two/three operators in tandem.
Rose explains: “The cost of building foundation models is expensive. So you want to find the best way to do that, and you can use the open source market, or go to the closed market; but there is a third middle-market, too, around this notion of a data trust. So if you build a data trust, between a few local operators within non-competing local markets, then [they can collaborate to] build and tune their own foundation models.” It is the spirit of co-creation, which, arguably, rival tech supply chains are better at, notably as complex Industry 4.0 problem-solving has improved.
The GSMA is handling the syllabus, in terms of the training, with IBM in support; IBM is providing the teaching materials, in the form of its watsonX AI and data platform, as well as the classrooms at its offices in Dubai, London, Mexico, New York and Seoul. Rose explains the logic: “There are a couple of problems that telcos still need to solve. One is the return on invested capital in 5G. There are significant initiatives to help with that, like the exposure of network functions across different APIs, and a number of forums to drive it, led by the GSMA, and TM Forum, too.”
He goes on: “But all of that takes a while. And the only way 5G will be consumed is via an ecosystem-play – via edge-based applications… Because access technologies alone aren’t good enough. And AI, as it applies across industries, will be one of the first options they can reach for. But for [AI applications], they need networks that are hybrid-by-design – which means hybrid cloud, hybrid data, hybrid AI. It is the constellation of those technologies that will enable 5G to grow, much more than the access technology alone.”
Those are the watchwords, it seems: network exposure, industry collaboration, solution building, value creation. But hang on; just explain this hybrid-in-triplicate philosophy? Because every technology is sold in hybrid form these days. What is hybrid data, for instance? Rose responds: “Yes, so this disaggregation of network functions has to be coupled to hybrid AI, hybrid cloud, and hybrid data – in order for the service to be delivered on three levels: with a meaningful TCO, with limited impact to the environment, and with differentiated SLAs for different industries.”
The other point, alluded to above in the references to tier-two/three carriers, is about the digital divide in the telecoms space – between both established and developing telecoms markets, and between both established and developing telecoms operators in them – which is liable to be cleaved open by AI if the technology is not understood and democratised. “AI acceleration is going to the tier-ones, with the most money, which can lean in very heavily to understand how to adopt AI,” says Rose.
“So you don’t just have a problem of 5G networks being deployed in leading markets first, but an even wider material gap potentially with AI – to the second and third players in the market, and with emerging markets.” So does he expect Vodafone and AT&T to rock up? No, probably not. But I am expecting tier-twos and -threes to rock up because, typically – and unfortunately, this is just the way the world works – the tier-ones get a lot of the attention. We want to drive equitability in the market.”
The press statement about the initiative included a bunch of stats, from IBM and the GSMA variously: 40 percent of telecoms is exploring or experimenting with generative AI, says one, and 56 percent of operators are actively trialling generative AI solutions, say the other; all of which means generative AI is being given higher priority versus other tech, they conclude, but that adoption is invariably less prevalent amongst smaller operators. The operator market is also behind most other enterprise sectors, suggests Rose – certainly in terms of urgency and appetite.
“The AI gap between telcos and enterprises cannot be allowed to grow either. Those enterprises are already moving at 1,000 miles per hour. There has never been a time in the last goodness-knows – maybe since the PC age – when all industries galvanised around one technology at the same time. So telco operators have seen the light; they have to move with this. It is not that they have a choice to slow this, to go at their own pace. The pace is already set; it is up to them to keep up.” The other race, which telcos are also losing, is the team sport of digital change – around co-creation, cooperation, or collaboration, or whatever it is called.
Rose says: “Yes, but the telecoms industry is learning at speed. It was slow because when it started to talk about things like open RAN there was a certain defensiveness with that learning process. But you have seen some break away from that, and AI has moved at such a pace that everyone recognises you have to collaborate on it. Because it’s simply moving too fast to solve all the problems or understand what’s going on in all these different domains. So the only way to solve for that is to become a better ecosystem, and mature your leadership accordingly.”
He shifts the angle, to explain again how this AI crutch will prop-up telcos, including as they interface with enterprises. “This is like the third leg of the stool. The first leg is the operator with its spectrum and channel to market; the second is the vendor with the technology; and the third is the hybrid cloud with its AI. Which is how telecoms goes from monolithic structures to hybrid structures, and from monolithic services to dynamic services – and how it serves multiple enterprise industries. It is a critical phase of enablement and democratisation, not just of AI, but all the parts of this three-legged stool – which make up the entire way telcos get to market.”
