• AI-driven network automation efforts captured the week’s main AI-native telco news headlines 
• Specification developments, the competitive threat of AI agents, and AI-related telco revenues were also notable themes 

Operators and vendors pressed ahead with AI-driven network automation this week, with major deployments and trials spanning the US, Bahrain, Thailand and Ethiopia, all while questions are being raised about the trustworthiness of AI agents. 

Standards body ETSI was also particularly active, pushing forward on AI specifications covering everything from deepfakes to 6G semantic communications. Meanwhile, BT’s CEO argued that telcos stand to gain from the agentic AI era, even as her company continued to shrink its workforce at a significant pace. 

So here are the main developments from the seven days that ended on Friday 6 November…

Major US telco Verizon is to deploy the Ericsson Intelligent Controller (EIC) nationwide as part of its ongoing work with the Swedish vendor to boost radio access network (RAN) performance and efficiency. The rollout of the EIC will support the deployment of AI-native rApps, which are modular, cloud-native software programs that can help automate, manage and optimise the RAN. Having already used its initial EIC deployment to run an in-house developed rApp that eliminates manual intervention from fixed wireless access (FWA) inter-radio access technology handover processes, the US telco now plans to deploy the EIC nationally this year, and will also test and deploy new rApps designed both internally and by Ericsson, including the Cell Anomaly Detector rApp, Root Cause Explainer rApp and Anomaly General Optimizer rApp, which aim to boost operational efficiencies. “Our expanded collaboration with Ericsson brings AI-native RAN automation that elevates performance across mobility and fixed wireless access,” says Matt Butz, VP of network planning at Verizon. “Using the Ericsson Intelligent Controller, and Ericsson rApps, in a live environment, we demonstrated AI cell shaping of massive MIMO and traffic steering with significant gains in throughput and efficiency. Verizon’s ability to self-develop rApps on this open platform will help usher in a new level of automation as we prepare for 6G,” he added. The deployment of rApps is a major step towards Verizon’s stated goal to reach Level 4 cognitive automation on its network, which the operator outlined in a blog last year. The EIC is built on open standards and is part of the vendor’s Intelligent Automation Platform, which it claims supports an ecosystem of more than 100 rApps developed by more than 100 members. “We’re proud to deepen our collaboration with Verizon to advance how AI transforms mobile networks,” stated Ravi Vaidyanathan, VP and head of cloud software and services for the customer unit Verizon at Ericsson Americas. “Together, we’re driving major gains in network performance and building a smarter, more efficient and autonomous network, unlocking better customer experiences and transforming operations.”

Bahrain’s mobile operator, Batelco, has partnered with Ericsson to expand 5G and introduce artificial intelligence capabilities into its network, leveraging the Swedish vendor’s radio access network (RAN) and core network solutions. The collaboration will see Ericsson’s RAN and cloud-native core technologies used to boost coverage, capacity and performance on Batelco’s 5G network in the Kingdom of Bahrain. The two will also work together to trial Ericsson’s AI-native Scheduler for Link Adaptation, which is an AI solution designed to boost network efficiency. Batelco, which is part of the Beyon Group, will deploy Ericsson’s massive MIMO radios to improve its downlink performance and energy efficiency, while also upgrading its packet core and unified data management system. Batelco also said it will move its cloud infrastructure from network functions virtualisation infrastructure (NFVi) to Ericsson Cloud Native Infrastructure Solution, which it hopes will help it to manage growing data demand in the Middle Eastern kingdom. Maitham Abdulla, Batelco by Beyon CEO, commented: “Our continued investment in expanding and modernising our network reflects our commitment to delivering an exceptional connectivity experience across Bahrain. By enhancing coverage and capacity while introducing advanced technologies, we are building a more intelligent and efficient network that meets the evolving needs of consumers and businesses.”

True Corp. is engineering and optimising its 5G network using AI with “the concept of Network Digital Twin, a technology that creates a virtual representation of the physical telecommunications network to simulate scenarios, assess potential impacts and identify opportunities for network improvement before physical deployment,” the Thailand telco noted in this press release. “This initiative lays the foundation for the future development of an autonomous network, with the aim of improving network efficiency and delivering better connectivity experiences to customers nationwide,” added the telco, which has 48.8 million mobile customers, of which 20.5 million are 5G users. True’s chief network officer, Khurrum Ashfaque, stated: “Our goal is not simply to expand 5G coverage but to develop a network that responds more effectively to how customers use connectivity and supports their constantly evolving lifestyles. Today, customers are consuming significantly more data, while different locations have different connectivity needs. True is, therefore, focused on enhancing network speed, capacity, reliability and management by combining advanced 5G technologies with data analytics and AI to deliver better connectivity experiences to customers across Thailand.

Ethio Telecom has partnered with ZTE to deploy Ethiopia’s first pre-commercial AIR RAN solution. Initial testing found the solution delivered an 11% increase in 4G capacity on the Ethiopian operator’s network, as well as a 40% gain in user throughput during peak hours. The solution uses frequency division duplex (FDD) massive MIMO technology alongside ZTE’s AIREngine computing solution to introduce AI-driven beam scheduling, which controls resource allocation on radio access networks in response to changing network traffic demands. It builds on two years of collaboration between the African operator and the Chinese vendor, ZTE said in a press release, during which they have tested how to improve RAN efficiency through AI solutions.

AI agents are proving rather more difficult to keep in line than their creators might have hoped. Recent incidents involving agents taking unexpected actions – including an AI system accessing an Australian government portal after its initial request had been denied – have put renewed focus on the risks of giving AI systems greater autonomy. The industry is moving towards increasingly autonomous networks, with AI agents being deployed to analyse network data, identify faults and recommend actions, while the longer-term ambition is for those systems to make and execute operational decisions without human intervention. For telecom operators, this is rapidly becoming a defining question: How much autonomy can operators safely give an AI agent if they cannot prove what it did, why it did it, or whether it remained within the boundaries they had set? This is the emerging challenge around trusted AI and autonomy – not simply whether an AI model can perform a task but whether an operator can assure its behaviour once it is given the authority to act. Read the full story.

Industry specifications body ETSI has been busy. It has teamed up with the ICT standardisation organisations from the Republic of Korea (the Telecommunications Technology Association) and Japan (the Telecommunication Technology Committee) to “explore the creation of a new global, industry-led partnership for AI standardisation” with the aim of forming an AI Partnership Project (AIPP), it announced. Bettina Funk, chair of the ETSI General Assembly, stated: “Global cooperation is essential in AI standardisation, even if priorities and expectations are not always the same. Exploring a new partnership with Japan and South Korea offers an opportunity to ETSI members to make progress on practical standards that respond to real needs. Such cooperation could bring valuable new perspectives and complement existing standardisation efforts, with a focus on delivering practical results.”

Separately, ETSI has released a new technical specification that aims to reduce the risk of generative AI systems producing harmful deepfakes. Developed by ETSI’s Technical Committee on Securing Artificial Intelligence (TC SAI), ETSI TS 104 275 “aims to strengthen accountability among AI developers and operators by providing practical measures and safeguards that demonstrate how to mitigate deepfake-related risks.” noted ETSI in this press release.

ETSI has also established a new Industry Specification Group on Semantic Technologies in AI-Native Networks (ISG SEM), “bringing together industry and research organisations to advance semantic communications and lay the foundations for future 6G standardisation,” it announced. The ISG’s mission is to “coordinate research into semantic technologies, establish common technical foundations and support their future standardisation as the industry continues to develop 6G networks.” Emilio Calvanese Strinati, convenor of the ETSI ISG SEM, stated: “As we move towards 6G, networks will need to do more than simply transmit data. They will need to understand and prioritise the information that matters, enabling more intelligent, efficient and responsive communications. But realising this potential requires a common approach across the industry. The ISG SEM will bring together valuable expertise from across the research and technology communities, laying the foundations for interoperable, AI-native networks.”

BT Group has published its financial results and operating update for the first six months of its current financial year, and there was a lot to digest behind the headline revenue, earnings and network rollout numbers. During the operator’s H1 financial earnings call held on Thursday morning, the UK national telco’s CEO, Allison Kirkby, made it clear she believes BT can benefit in the agentic AI era. Responding to a question from Goldman Sachs analyst Andrew Lee, who asked whether the use of AI agents could prove to be negative for telco revenue trends and whether the savings from using AI would offset the impact of agentic AI-enabled pressures, Kirkby noted: “I still believe telcos, like BT in particular, will be AI winners, because AI has dramatic potential to improve how we serve our customers, how we run our operations, and how we just move faster in everything we do. What are we seeing? AI is one part of a multifaceted transformation, business transformation, both cultural and operational at BT. We are kind of embedding it everywhere.” She noted that while the use of AI was not yet making its mark on BT’s financial performance (in terms of direct lower costs), the benefits can already be seen in how call centre staff are “able to use it for call summarisation to get better insights and information on their customers so they can do better cross-sell and upsell. We are seeing real benefits in our B2B division, where we had a lot of legacy products and services and very poor visibility on end-to-end processes.” Kirkby noted that BT’s team “created an agent recently to improve the bids, the complex bids, for some of our more complex customers. That agent has already reduced the bid time from 14 man-weeks to three man-days. Just think about the scaling opportunity of that. We remain convinced that AI will have great operational and customer service improvement.” And in terms of the threat from agentic AI, she noted that “we need to start feeding the models and feeding the agents with all of the great knowledge, not just about price but the quality services and the benefits of convergence of all of our brands. The agents are this new channel. They understand the true value of buying a product and service from a BT brand versus others… Overall, AI will definitely be beneficiary to a business like BT as it continues on its transformation.”

Headcount reductions are front of mind in the UK telco sector right now, following Virgin Media O2’s announcement earlier this week that it is set to cut its staffing costs by 30% by shrinking its workforce. That process has been underway at BT for some time already and, as part of its (revised) plan to reduce its annual operating expenses by £3.7bn by the end of March 2030 (the end of its FY30), the telco had always targeted significant cuts to its total labour force (TLR), which comprises “direct labour” (salaried staff) and sub-contractors. In its H1 financial report, BT noted that it had reduced its TLR by 11% during the six months from April to September, from 102,000 to 90,000 people. It also noted that its TLR target for the end of March 2030 is now in the range of 65,000 to 70,000, lower than the 75,000 to 80,000 range it specified only six months ago, though this is mainly because the numbers exclude the BT International workforce (about 8,000), which is being offloaded into a joint venture with Verizon that was announced in June. If BT ends March 2030 with 65,000 TLR, it will have reduced its workforce by 55,000 (almost 46%) in the space of just six years as a result of redundancies, portfolio changes (divestments) and a reduced need for contractors once the fibre-to-the-premises (FTTP) rollout is concluded. Whatever the process, that is a massive drop in human resources, part of which will be related to the increased use of AI for process automation. Of course, BT could be adding additional staff if it gets the regulatory green light for its planned and somewhat contentious acquisition of broadband service provider TalkTalk, but Kirkby was not able to comment on that M&A deal today due to the ongoing review of the planned transaction.

SK Telecom (SKT), which has long described itself as an AI company, has reported its third-quarter results: The update includes figures for its AI-related revenues, which are growing, but not at breakneck speed. Total group revenues increased by 13.1% year on year to almost 4.5tn won ($3.36bn) “driven by growth in the company’s wireless and fixed-line telecommunications businesses”, while operating profit grew significantly to 552.3bn won ($410m) compared with the paltry profits reported last year when SKT was still reeling from a major cybersecurity incident. SKT’s AI datacentre (AIDC) unit reported third-quarter revenues of 126.1bn won ($94m), up by just 6.2% year on year, while the company’s consumer and enterprise AI service revenues grew by 27.3% to 70bn won ($52m). The operator noted that it is “on track in its AI services business. The company is expanding the use of its proprietary AI foundation model across its A. (A-DoT) AI assistant and workplace AI services. Through its participation in Korea’s ‘AI for All’ initiative, SKT aims to combine its telecommunications and AI capabilities to build the nation’s largest AI services ecosystem,” it added. The operator will be looking forward to much bigger returns in the future as its AI businesses matures: In August, SKT announced the spin off part of its SK Broadband subsidiary to form a new company, SK Horizon, that will focus on the management and development of AI datacentres in South Korea; and in July it formed its AI factory unit, SK Hyper, which will develop gigawatt-scale projects. On the company’s earnings call, CFO Park Jong-seok explained how SKT sees the telecom and AI parts of the company becoming integrated. “As AI technology advances and its range of applications expands, the relationship between telecommunications and AI is becoming even closer. Against this backdrop, we aim to implement a single growth strategy for both telecom business competitiveness and AI business development. We will apply AI technology to strengthen telecom services tailored to customers and improve network operation. While utilising our telecom customer base and network capabilities to expand the adoption of AI services across everyday life and industries, we will develop AI infrastructure not only as a foundation for AI technology but also as a key growth driver in the medium and long run. We will ensure that these achievements we have made in our telecom and AI domains will lead to bigger growth.” SKT ended September with 21.9 million mobile customers (of which 18.1 million are 5G users) and almost 7.4 million fixed broadband customers.

Deutsche Telekom, which counts numerous operations across Europe as well as T-Mobile US in its corporate portfolio, continues to grow in strength, reporting third-quarter revenues of €31.1bn, up by 5.5% year on year on an organic basis (excluding the impact of exchange rate changes or changes in corporate composition). Organic service revenues grew by 4.1% to €26bn, while organic EBITDA after leases increased by 8% to €12.2bn. CEO Tim Höttges noted: “Our strong figures are setting benchmarks in our industry. The consistently impressive results we have delivered for many years now bear testimony to the success of our strategy.” The operator noted it was seeing strong demand at its T-Systems enterprise services division (referred to in its financial reports as Systems Solutions) for its Industrial AI Cloud services: Last month DT noted during its first ever AI Investor Day that it expects to generate an additional €1bn-plus in revenues outside the US from AI-related services by 2030.

Meanwhile, T-Systems is adding AI infrastructure company Sparxity (formerly SpiNNcloud Systems GmbH) to its AI partner ecosystem: The two companies aim to combine Sparxity’s “sparse-native computing technology” with T-Systems’ datacentre infrastructure and operational expertise. According to DT, the two companies will “jointly develop and deploy the next generation of energy-efficient AI computing infrastructure in Europe. Sparxity’s technologies are used for so-called energy-saving and demand-driven ‘sparse computation’,” which involves “directing computing power exactly where it is needed”. Initially, Sparxity’s current hardware will be used to jointly develop approaches for the technical integration of the next generation of Sparxity chips with T-Systems’ datacentre infrastructure and partner ecosystem, with the first deployment targeted in 2028. “Building on this, the two companies plan to develop high-performance, energy-efficient AI computing systems for future applications and optimise them for specific use cases,” added DT. Ferri Abolhassan, CEO of T-Systems, stated: “Germany and Europe need a powerful AI infrastructure that can be operated and scaled economically. With our T Cloud portfolio and our AI factory in Munich, we have created the basis for this. We are open to further partners and are looking forward to having Sparxity on board in our AI factory soon. The German company scores particularly well with its energy-efficient use of AI computing power.” And in case it was left unsaid, Karsten Wildberger, Germany’s federal minister for digital transformation and government modernisation, hammered home the sovereignty message: “The project shows that digital sovereignty is feasible in Germany – if we combine our own technological expertise with efficient infrastructure and strong partners.”

Orange has partnered with Algolia to deploy AI agents into regulated customer support conversations, with pre-selected guardrails defining what the agents can discuss. The French telco will use Algolia’s Agent Studio across its business-to-consumer (B2C) and business-to-business (B2B) support operations in more than 26 countries. The rollout builds on Orange’s existing Algolia search technology across 24 markets, where deployments have delivered a 63% increase in order rates. Algolia even shared some numbers, saying its search system had resulted in a more than 63% order rate, 14% click-through rate from its neural search system, and +7 points in net promoter score (NPS) for Orange’s customer support. Orange Group started with Algolia InstantSearch, then added further capabilities, including filters and facets, AI-driven synonyms, dynamic re-ranking, collections, query categorisation, AI recommendations and advanced personalisation. Its Agent Studio extends this into conversational and agentic experiences. Nicolas Coudert, digital factory director at Orange Group, said: “An effective search engine should do more than return a list of pages, it also should provide a direct and relevant answer. In a complex digital environment, responding instantly improves the customer experience, which drives higher satisfaction and stronger engagement. Algolia pairs reliable, high-performance services with a steady cadence of innovation, such as Agent Studio, that we can integrate quickly and put in front of customers immediately.”

- The staff, TelecomTV