Why mobile operators need AI
by Baruch Pinto | September 17, 2025
Earlier this year, we discussed the role of cellular connectivity for AI in IoT applications across different verticals. Now is the time to flip the perspective and look at how important AI is for cellular connectivity itself. We briefly touched on it in our biannual Chronicles, let’s explore it in more detail.
Why do mobile operators need AI?
The number of cellular IoT connections reached 4 billion at the end of 2024 and is forecast to exceed 7 billion by 2030. All these endpoints generate a huge and constantly growing amount of data. Juniper Research estimated it at 21 PB in 2024 and expects it to more than double by 2028. On top of that, this is mixed traffic that includes periodic meter pings, bursty alarms, firmware updates, or time-critical control data.
To make sure that a utility meter, a video surveillance camera, and a connected car can share the same network and each get the performance it actually needs, mobile network operators must move beyond raw connectivity. Since different devices and applications have specific connectivity requirements, “best effort” isn’t enough, and telcos have to guarantee outcomes such as latency, jitter, loss and reliability. They also need to be application-aware, classifying traffic and steering it to the right local breakout so the workload runs close to the device.
At this scale, manual-only NOC workflows can’t keep up, and that’s where AI and machine learning become essential for forecasting, real-time anomaly detection, policy optimization and control. Most carriers have already implemented it in some form: according to a 2024 study by Nvidia, nearly 90% of telecom companies use AI, with 48% in the piloting phase and 41% actively deploying AI. Most telecom service providers (53%) agreed or strongly agreed that adopting AI would provide a competitive advantage.
With the level of adoption so high and so many AI applications available today, how exactly do telcos use AI? There are several layers where MNOs and MVNOs utilize AI to optimize their operations, increase uptime and reliability, improve customer service quality and cut costs:
Network Management Automation Layer
AI enables dynamic adjustments to handle network traffic. In networks that include SON capabilities and automation features, it helps the network to auto-configure and self-optimize in real time. AI tools can monitor network performance and automatically adjust parameters like signal strength or bandwidth allocation. They can also detect and mitigate issues like cell outages or traffic congestion without human intervention. For instance, AI can identify a congested cell tower and reroute traffic to neighboring cells to maintain service quality.
There are real-life examples of it: Norway-based Telenor has integrated AI for network configuration, utilizing agentic AI to autonomously optimize network parameters. This system dynamically balances factors like speed, interference, and energy efficiency.
Another use case for AI can be analytics that help steer outbound roamers to preferred visited networks for quality and cost. There seems to be a clear shift toward AI-assisted steering of roaming in the industry: Kaleido Intelligence Q3 2025 survey shows that nearly 4 in 10 operators (39%) plan to adopt real-time, quality-based steering within the next year, while industry players discuss “adaptive” and “real-time intelligent” steering or explicitly write about building AI into steering of roaming.
Network Maintenance Layer
Using AI models, operators can analyze data from network equipment, from traffic flows and error rates to temperature or power levels. This helps predict failures before they happen and schedule proactive maintenance to reduce downtime and improve reliability. Also, for network operators, this reduces OPEX by minimizing emergency repairs and extending the life of assets.
Preventive maintenance goes beyond just monitoring the hardware or software health of the network, it can also involve proactively managing network load. For instance, if there’s a sudden traffic surge in a specific area due to an accident, AI can react faster than a human operator to mitigate potential issues.
According to experts’ insights, operators can greatly benefit from using autonomous digital agents powered by generative AI, for maintenance purposes. These bots or systems can be embedded in the core of back-end operations and interact independently, handling large workloads. For instance, they can conduct root cause analysis on performance degradation and resolution. Implementation of AI-driven maintenance operating procedures can reduce maintenance costs by 25%-30% and provide predictive demand and capacity recommendations that lead to 5%-8% optimization in CAPEX. Olyver Wyman estimates that it can also improve efficiency of NOC, reducing required staffing level by 2-3x.
Many operators are already using AI for predictive maintenance and improving efficiency of their networks in terms of power consumption, or performance optimization. For example, Telefónica Deutschland is using an AI tool to predict mobile data usage patterns for more precise capacity planning, targeted investments and optimized network expansion. The AI analyzes data from 28,000 radio access network sites and factors in other information, such as mobile package tariff changes. According to the operator, the system provides twice as many accurate predictions as conventional planning tools and has an accuracy rate of more than 90%.
Another important application of AI is green telecom initiatives. AI can optimize power usage in base stations and data centers by dynamically switching off underutilized equipment or balancing load based on energy consumption models. In March, Vodafone UK informed that it managed to reduce daily power consumption of its 5G radio units by up to 33% at select sites across London, without impacting the service their customers receive.
Customer Service Layer
Customer service is probably where AI has been implemented the most so far. Both MNOs and MVNOs use AI extensively to improve customer experience and optimize support. They deploy AI-driven chatbots and virtual assistants to provide 24/7 help, and use predictive analytics to help tailor offers and anticipate customer needs. For example, AI-powered agents in call centers can suggest solutions by analyzing customer intent and history.
There are plenty of real-life use cases.
Last year, Verizon reported that it was using generative AI to stop 100,000 customers from leaving its service by predicting why a customer is calling, connecting them with a suitable agent and reducing store visit time by 7 minutes.
Optus, the second largest telecommunications company in Australia, uses AI to identify faults and customers’ issues, enabling them to simply solve their own problems, while enabling the operator to look at customer segmentation in a more granular way, to deliver better offers and products.
BT enhanced customers’ experience with AI, helping them prepare for international travel. Eventually, the company halved the need for online chat and messaging support, thanks to AI’s ability to understand and respond to customer needs. It is also used for billing support, where generative AI provides detailed explanations of billing charges, enhancing transparency and customer satisfaction.
McKinsey reported that a leading European telco deployed an AI-driven help desk bot that led to a 35 percent reduction in cost per call and a 60 percent higher customer resolution rate.
Mass adoption of AI by telecommunication companies is not surprising, because the benefits in cost and resource savings are clear, but it’s not just operators who benefit from AI. Enterprises get more predictable, SLA-backed performance for their IoT fleets. Device manufacturers and IoT providers see simpler onboarding and lower data/energy costs. Even for end users, AI ensures fewer outages and quicker support.
However, operators now must answer technical questions: where to run AI (core, edge, or device), how to govern and scale it. Placement can be layered depending on the use case. In the core or in the cloud, AI can handle capacity planning, roaming steering based on cost and quality, or generative AI for customer care and operations. At the edge, near-real-time models can deliver RAN energy saving and load balancing, application-aware traffic steering, and local breakout for low latency.
With so many capabilities that AI can offer, the pragmatic path for the operators that haven’t implemented it yet is to start with a few use cases, be it energy optimization or predictive maintenance, run focused pilots, measure what improves and scale what works. For the operators that are already getting results from AI, the next step is to widen the impact by rolling their AI tools out to more markets and domains, keep it up to date and tie outcomes to their business KPIs such as uptime or cost-to-serve.
As a global MVNO with a full core network, Webbing has many applications for AI in all of the layers we have discussed, from steering roaming traffic or packet gateways load balancing to customer support. Besides, AI tools help us to perform business analytics, speed up onboarding and ensure quicker issue diagnosis and resolution. We’re working on expanding AI across our workflows to further improve efficiency and performance.
