MWC26 ON TREND SPECIAL
Is there method behind the AI madness?
COMMENT
Is there method
behind the AI madness?
The story at MWC26 Barcelona was again about AI but this year there was less hype and more realism.
That’s not to say AI has arrived in telecoms fully mature and everyone has a clear vision of their AI-fuelled future. In fact, the honest people were quick to admit incomplete understanding of where AI will take the industry and they were keen to point out that it’s still very early in terms of adopting the technology beyond simple pilot projects.
The show itself seemed buoyant in spite of the onset of war in the Middle East. This prevented many visitors from the region and people routing through its transit hubs from attending the show and it was noticeable that fewer delegates were roaming the halls, certainly on day one. The organisers claim 105,000 visitors and this should be no concern for an event that has proven its resilience over the last 20 years, which included the 2008 financial crash and the pandemic.
For telecoms – and this year’s show clearly transcended mobile communications – there’s plenty of innovation and another round of complex challenges to address. From my seat, I can’t help but wonder whether AI in telecoms today is a set of technologies looking for a process to make autonomous. Autonomous networks, automated operations and connected experiences will all utilise AI but these are emerging rather than emerged technologies.
I’m confident that mainstream, large-scale telco use cases for AI will become apparent but it’s above my paygrade to make a call on when or how they will be implemented.
INTERVIEW
"We support AI and automation efforts so we can implement an autonomous system so our life will be easier and we will maintain the same service with less effort."
Kazuhiro Furuhata
chief network officer,
KDDI Corporation
Kazuhiro Furuhata, the chief network officer at KDDI Corporation, met Agile Telco’s George Malim at MWC26 Barcelona to discuss AI, automation and the road ahead for efficient operators
For many telcos, AI opportunities come in two forms. There are the as-yet-unrealised revenues from new services enabled by AI and the soon-to-be-generated efficiencies and operational gains that AI unlocks in the network. KDDI Corporation has both bases covered with a booth at the event showcasing autonomous driving, humanoids and partners that are already utilising AI. It’s still early days for the technology, acknowledges Furuhata-san, who is also an executive vice president at the telco. “There is not much [pressure on the network] at the moment but latency is going to be very important for AI,” he predicts.
The more immediate priority is to use AI to enhance automation and achieve efficiencies in network operations and Furuhata-san confirms KDDI’s aspiration to achieve Level 4 of TM Forum’s Autonomous Networks Maturity Model. “Within our operations we have started to use AI, for example, to determine the root cause of issues and any interruptions,” he adds. “We also see use cases for base station perimeter optimisation.”
AI augments operations
There is potential for these types of systems to be applied to additional use cases. Furuhata-san cites the example of earthquakes which occur regularly in Japan and can disrupt both wireless and fixed line networks. AI can be utilised to provide an early warning, accelerated identification of a root cause and to help model optimal responses such as apportioning greater wireless capacity to an affected area.
“Of course, the network is becoming more complex so, for example, if an earthquake cuts an optical cable we can have a faster response,” he says. “A lot of facilities have alarms already but if there is AI monitoring the network, it will be very early to determine the root cause of the issue so that’s something we can help with.”
Furuhata-san also expects improved monitoring, automation and autonomy to help smooth demand for capacity in the wireless network. KDDI operates both 4G and 5G networks and sees both as effective ways to meet demand and control availability of bandwidth. “Depending on the area, it might be better to use 5G but sometimes it’s better to use 4G,” he says. “It depends on what type of data needs to be transmitted and how much data is in the queue.”
Automation and autonomy
“As a network provider, we operate the telecoms infrastructure so the workforce we can have is large,” Furuhata-san explains. “We support AI and automation efforts so we can implement an autonomous system so our life will be easier and we will maintain the same service with less effort.”
That starts with having the correct data to enable automation. “Data on human contact is chaotic and it’s not appropriate to use for organisations so the first step is to clean the data and that’s something we’re working on,” he says. “With machine-to-machine communications, transactions will multiply and when people speak to each other there is a lot of data. This means there will be a lot of data and transactions so we need a network that can accelerate that. We believe it might be better to separate the smartphone and the machine networks in the future. Until now, smartphones have just sent data vertically but machines will involve horizontal traffic as well.”
Achieving that acceleration is all part of KDDI’s strategy. “We do have a business plan for this scale and also achieving a return on investment from making investments based on these trends,” Furuhata-san concludes. “Japanese telcos continue to discuss in-house how to invest efficiently.”
EVENT REPORT
MWC26 Barcelona - Is there method behind the AI madness?
Sometimes good things are worth waiting for. Co-incidentally, in the run up to MWC26 Barcelona, builders completed a glass cross on the top of the Sagrada Familia basilica in the city. That move brings the building to its final height – just 144 years after work began. AI adoption in telecoms will certainly take less time but it hasn’t been immediate as the gap between last year’s hype and this year’s reality has shown.
Last year, the sentiment was that AI can be a universal panacea for telecoms, enabling telcos to simultaneously automate their operations, radically control their cost base and give their customers autonomy to take what they need from the network automatically. Telcos would be living a dream in which their newly automated infrastructure runs lean and cheap while they pile on margin for connecting AI. The reality is we’re somewhere between the extremes, nearer the start than the completion of the vision.
Telcos are using AI to make their network operations more efficient and to automate some functions but these remain discrete. There are methodical steps for the industry to go through to prepare for internal usage of AI and the presentment externally of AI-enabled services and infrastructure. MWC26 Barcelona revealed that method is now outweighing madness with careful approaches being adopted behind the scenes to create the foundation for AI in telcos. The jazz of robots at booths and far-fetched use cases is still present but these are there more to entertain the delegates than to actually sell a service or product.
Sophistication has to wait
Sophistication will come later and few see telcos taking steps to rely on AI-enabled automation of critical systems in the short-term. Instead, AI and agentic AI are being used this year to pick around the edges. Additional call centre automation and some preventative maintenance use cases are the real-world examples right now. The robot dogs will be consigned to kennels for the foreseeable future.
This isn’t to suggest AI won’t happen for telcos. Far from it, it’s just that the technology is immature, developing fast and untried in the high-reliability, heavily-regulated telecoms industry. What is needed, who will be partnered with and how AI tools and systems will be deployed are still being decided and most recognise this is complex and rapidly changing.
Yesterday’s AI for today’s networks
“AI is on everyone’s lips but lots of people, including myself, don’t necessarily understand it,” observes Ultan Mulligan, the chief service officer at ETSI. “Operators are making genuine use of AI around network monitoring, configuration management and optimisation with real results on networks that were never designed with AI in mind. In these scenario’s yesterday’s AI can generate results.”
Mulligan is looking ahead to networks that are designed with AI in mind and that will involve 6G which he doesn’t see being deployed until 2030-31. “AI is going to be part of 6G but that is in its infancy and the agentic AI-driven network is not here yet,” he explains. “Operators are trying to embrace hyperscalers’ culture quickly and have strongly embraced continuous innovation, continuous development (CI/CD) pipelines for speed to market of new functions and features. Even without AI this is a really significant embrace of CI/CD processes in the telecoms space.”
AI-natives
He’s not alone in acknowledging the pace of application of AI in telecoms has been overstated. John Harrington, the executive vice president and head of Europe at Nokia, says: “Mention AI and people glaze over through a lack of understanding but for us it’s really about connecting the intelligence and to go a bit deeper, we’re building AI-native networks.”
“We used to connect people and now we connect intelligence, every generation has had a dominant workload of the network,” he explains. “First it was voice, then messaging and data, then video, then M2M and now it’s tokens – a bunch of data. The traffic pattern and the workload of AI is changing the demand on the network.”
“AI is not just a feature, it’s a different structural workload so you need a new architecture,” he confirms. “AI-native networks need an architectural shift.”
That shift is underway but comes with mistakes, over-ambition and project failures. “Last year at MWC, the hype was about AI,” confirms Ilan Sade, the head of the GenAI and Data Division at Amdocs, which launched it aOS agentic AI platform just before MWC26 Barcelona. “This year we are starting to see real things happening. Enterprises in general, and telcos are no different, are not doing it at scale. Proofs of concept are failing or not getting to scale and that’s what’s stopping the telecoms industry engaging the technology.”
head of GenAI and Data Division,
Amdocs
Sade emphasises that telcos face the additional challenge of operating telco-grade environments and this demands accuracy and precision that AI lacks. “The industry has to go back to basics to understand what it means to deploy a technology in that environment,” he explains. “They need an enterprise AI infrastructure ecosystem that they control and they need to host solutions, some of which they develop themselves and some from others.”
Agents of change
Agentic AI was certainly on almost everyone’s lips and significant benefits are anticipated but telcos are keen to inject some caution into the excitement. “We will do AI and put agentic automation in place but not in the delivery chain because every code push coming in is going to do two things: an agent is going to review the code and parse for configuration and then it will not allow the change,” explains Subha Shrinavasan, a senior vice president of Global Services at Rakuten Cloud. “AI will tell the agent what will go wrong and AI agents will be told to continue to watch out for issues.”
senior vice president of Global Services,
Rakuten Cloud
“There’s a lot more to optimise in the delivery pipeline in AI right now and top of mind is what it could be,” she adds, pointing out that reality dictates seeking out winning use cases. “We are answerable to the c-suite so we need to show tangible results in terms of uptime or improved support processes. In the human contact centre people get asked the same questions so these are areas where we have AI agents which are reducing operating expenses by not requiring humans to answer questions where we already have information to answer them in the knowledge base. Customers like self-sufficiency and CSAT goes up when they are able to do it themselves.”
Shrinavasan emphasises that Rakuten Mobile is fully committed and has introduced a full-blown implementation of AIOps. “Today, AIOps is a framework and the only currency we have is the data and data from other businesses in our group,” she explains. “AIOps does data cleansing and organisation and plays a critical role using custom-built models with historical data to find anomalies and point to issues.”
For Sade at Amdocs, there’s a steep learning curve for telcos to climb. “With the agentic vision three, four or five years out, it’s clear that many things will be agentic but not how to get there,” he says. “You need a partner who understands what it means to be up and running and then have the ability to do it in a responsible way.”
Hurry cautiously
There’s an odd sense of time running out but also needing to be spent on ensuring AI works for telcos. The hyperscalers, enterprises and consumers see AI revolutionising businesses, processes and lives but the telcos need to not only adopt AI in their operations, they need to embed it in their telco-grade infrastructure. That can’t be rushed if it is to be achieved with the high quality needed.
MWC26 by the numbers
AI at MWC26
Visitors claimed by MWC26 organisers
0k
Operators prioritizing AI monetization as their top strategic business goal.
0%
Governmental delegations participated in the Ministerial Programme.
0
Exhibitors, sponsors, and partners
0
Attendees from adjacent industries
0%
Journalists and industry analysts reporting from the event
0
In contrast and at the same time, telcos realise they need to use AI to enable AI-related services and generation of AI-based revenues. The autonomous AI-augmented mission critical infrastructure hasn’t been built yet for the AI-enabled world to run on top of. Telcos alone can’t make that happen but they know they need to play their part if they are to have a future in which they are anything more than a commoditised network infrastructure provider.
The good news is that the vendor community is rising to the challenge and redefining how it supports telcos. Nokia, for example, has shifted its architecture – NVIDIA has invested US$1bn in 3% of Nokia and is collaborating on GPUs that are fit for purpose for AI-native networks. Those are different to traditional data centre GPUs and Harrington says that adding GPUs to cellular base stations is now a reality. In addition, Nokia has acquired Infinera to add data centre interconnect expertise to its portfolio.
Orchestration for a different network tune
“It’s a different Nokia and it’s quite tangible,” he says, pointing to enablement of automation and autonomous networks. “What do autonomous networks mean? You need more than an element manager for the RAN, you need orchestration over the top and then you add AI and it becomes agentic AI where the agent looks at the throughput and load steering. We’ve invested a lot in a very broad portfolio so we can compete in autonomous networking.”
“We have automation solutions in each of the radio, transport and core domains and we’re working with customers on the TM Forum Autonomous Networks Maturity Model – everyone wants to get to Level 4,” he adds. “Operators are trying to reduce costs, reduce opex and be more automated. That involves moving to agentic operations because in future you won’t have managed services, you’ll have agentic AI in operations.”
Method, not madness
This struggle to make effective use of yesterday’s AI in current networks while preparing for autonomous operations in next generation infrastructure demands a split personality within a telco. Innovation must be applied to current operations to optimise and allow the network to support customers’ needs for new services but, at the same time, investment must be devoted to the next generation.
Telcos have always done this, having the agility to operate legacy networks while building out the next generation of mobile or IP networks. That multi-directional strategy experience places them well for the shift to AI. They will be able to draw on that to methodically implement AI and maximise the benefits of the technology for users and in their own businesses but only if they carefully deploy, tested, robust and ultra-scalable platforms for automation and autonomy.
MWC26 Barcelona suggested that there is method, not madness, in the telecoms industry’s approach to engaging AI in its operations. It’s time to drop the droid demos and focus on creating essential AI enablement infrastructure.
EDITORS TAKE
Is there method behind the AI madness?
Have your say
Give us your opinion on AI in telecoms, automation and whether the industry is finally replacing hype with method. Ring in and leave a message or use email to comment.