From connected to intelligent: How AI is transforming Vodafone's network

AI is changing how networks are planned, operated and used. For customers, the goal is simple: connectivity that adapts to their needs, fewer service interruptions and the introduction of new services that make everyday life easier.

Imagine taking a journey from a busy shopping centre to your home. Along the way, you make a call, consult an AI assistant and join a video meeting. The technology behind the connection matters less than whether it delivers a seamless experience when needed. Vodafone’s network transformation is increasingly focused on that user experience. One that anticipates demand, adjusts resources dynamically and resolves issues before customers are affected.

The scale of the challenge is growing. Vodafone’s mobile data traffic has almost tripled in five years, while traffic on its pan-European fixed network has increased by almost 60% over four years. AI is transforming both sides of the connectivity equation: giving operators new ways to build and run networks while enabling a new generation of AI applications with growing network demands. The transformation is therefore not simply about adding AI to existing processes. It is about rethinking how the network operates, what it can offer and the role it plays in a world increasingly shaped by artificial intelligence.

Vodafone's vision is a network that not only provides a high-quality connection, but also one that anticipates demand, dynamically adapts to changing conditions and enables intelligence wherever customers need it. Three interconnected pillars are shaping this journey. First, AI for Networks transforms planning and operations. Second, AI in Networks embeds intelligence into network functions and services. And third, Networks for AI brings together connectivity, distributed computing and cloud platforms to deliver AI experiences. 

These are not separate destinations or sequential programmes. They are complementary transformations, moving the network towards greater autonomy and a broader role as a platform for innovation. To scale autonomous networks, Vodafone has launched a multi-year, company-wide programme that will progress from individual use cases to a fully autonomous, multi-domain network.

AI for Networks: from manual actions to fully autonomous outcomes

AI for Networks changes how the network is designed, operated and enhanced. The journey involves moving from reacting to events towards anticipating demand, as well as from managing individual assets towards maintaining the experience a customer depends on. Rather than treating capacity, faults and energy as separate problems, the ambition is to make decisions that collectively focus on service outcomes.

This shift in approach begins with a better understanding of the network. Demand forecasting and digital twins help engineers explore changes before making them. Once the network is running, performance data can support earlier diagnosis and more informed action. Vodafone’s work with digital-twin technology is an example of this direction. It involves mapping tens of millions of environmental features alongside network performance data to improve decisions in the physical network. 

Vodafone’s trial of an AI-controlled antenna brings that idea into the physical world, directing capacity towards a shopping centre by day and nearby homes in the evening, with adjustments taking around 20 to 30 minutes rather than days for a manual change. In the core network, optimisation of the User Plane Function (UPF), which is responsible for routing and directing all data traffic, leads to hardware and energy efficiency. The larger transformation is from isolated solutions to a continuous cycle of observation, analysis, decision and action.

Automation is an important foundation, but autonomy goes further. An automated workflow follows a defined sequence; an autonomous capability seeks to maintain a desired outcome as conditions change. People set objectives and boundaries, while systems determine and execute appropriate actions within those controls. In practice, Vodafone’s energy-intelligent use cases have lowered daily 5G radio power consumption by double digit percentages, and orchestrated upgrades to thousands of routers have been carried out almost 70% faster and with zero incidents. Vodafone is working towards the aim of reducing customer service interruptions by up to 50% and restore service 70% faster.

For the network, it means adapting resources without treating every change as a new manual task. This automation provides benefits across the spectrum, such as an annual reduction of tens of GWH of energy, and 25% fewer service issue tickets raised per month while at the same time guaranteeing much faster repair times. Furthermore, RAN optimisation eliminates poor quality of service to the user by reducing latency to just 10s of milliseconds and improving session continuity by several percentage points. 

Vodafone's approach combines coordination across network domains with specialised controllers closer to the radio, transport, fixed and core networks. This unites the broader service objective with the expertise needed to act locally. The journey builds on existing automation rather than replacing it entirely, scaling Level 3 capabilities and pursuing Level 4 where value and risk justify it. Autonomy is a progression of capabilities, not a single label applied to the entire network.

AI in Networks: from carrying services to enhancing them

AI in Networks changes what the infrastructure itself can do. Intelligence moves beyond tools used to manage the network and into its functions and services. The ambition is not simply to carry a call or establish a connection, but to make it more useful. Embedded models and suitable computing capabilities create opportunities for network functions to adapt and for familiar services to gain new capabilities. The transformation also reaches the radio itself, where embedded intelligence can improve how signals are understood and resources adapted, making the infrastructure more capable as well as the services it carries.

Vodafone’s demonstration of language translation and noise cancellation during calls, without an app or modified smartphone, offers a glimpse of that direction. The trial translated conversations across five languages, including English, Spanish, Italian, French and German. Its significance is the change in where intelligence sits. In other words, a service people already use can be enhanced inside the network, rather than relying solely on a new device or application. In the core network, optimisation of UPF leads to reduced delays in processing customer’s calls. The journey is towards infrastructure that is not only a delivery mechanism, but that contributes directly to shaping the experience.

Networks for AI: from transporting data to delivering intelligence

Networks for AI is about extending the transformation beyond the network itself. As people and devices increasingly interact with AI, connectivity and computing must work together to deliver the best end user experience. Cameras, smart glasses and agents can send images, video and content upstream, emphasising the importance of capacity on the upload link.  Even a relatively modest uptake of AI/AR smart glasses could see a total uplink traffic increase three times the level it was in 2025 by 2031.

Responsiveness also matters. For example, how quickly an AI based answer begins to be delivered (time to first token) and how smoothly it continues (tokens per second) are key measures in a world where tokens are becoming a clear metric and currency of workloads. Network transit, computing queues and model processing all contribute. Bandwidth alone is not the answer. A double-digit reduction in end-to-end network latency thanks to L4S technology enables much faster AI consumption, while new compression algorithms support faster video upload and download.

The journey therefore involves connecting intelligence across device, edge and cloud, with workloads placed according to latency, processing capacity, power and data locality. The nearest computing location is not automatically the best one. This is where operators have a broader role to play by combining connectivity and computing to support AI applications, rather than merely carrying their traffic.

Together, the three pillars describe one transformational journey. AI for Networks makes operations more adaptive. AI in Networks makes infrastructure and services more capable. Networks for AI provides connected computing foundations for new experiences. Progress in each strengthens the others, but trust must remain common to all. It is security, privacy, operational controls and human accountability that set the boundaries. 

Returning to the example of a journey from the shopping centre to home, the ambition is not for customers to notice more technology. It is for their connection to support what they need, wherever they need it. Vodafone’s direction of travel is from delivering data towards enabling outcomes. A network that increasingly anticipates, adapts and brings intelligence into everyday life.