Much of the discussion around 6G still gravitates towards peak data rates, new spectrum and futuristic use cases. Yet one of the most profound shifts promised by 6G is not about raw performance at all, but about how networks understand and respond to individual users. A recent Analysys Mason paper explores this idea in depth, arguing that 6G could enable true hyper-personalisation by embedding agentic AI throughout devices and the network itself.
Personalisation in telecoms is not new. Operators already differentiate services through network slicing, quality of service tiers and limited quality on demand offerings. However, these approaches are largely static, semi-personalised and constrained by the architecture of today’s 5G networks. They struggle to react in real time, lack deep awareness of user context and rely heavily on best-effort assumptions. As applications become more interactive, immersive and AI-driven, these limitations become increasingly visible to users.
Hyper-personalisation goes far beyond tailoring a tariff or boosting throughput for a predefined user group. It refers to the ability of the network to adapt continuously and in real time to an individual user’s situation, intent and experience. This could mean maintaining a consistent quality of experience as a user moves between environments, applications and devices, or proactively adjusting network behaviour before the user even notices a problem. Achieving this requires a step change in how networks sense, reason and act.
The paper makes the case that agentic AI is central to this shift. Agentic AI systems are designed to operate autonomously, making decisions and taking actions based on goals rather than explicit instructions. In a 6G context, this means AI agents embedded in user devices, referred to as agentic user equipment, working in tandem with AI agents in the network core, often described as the agentic core. Together, they form a closed-loop system that continuously exchanges intent, context and feedback.
One of the most interesting aspects of this vision is the role of the 6G core network. Rather than being a relatively passive control point, the core becomes the brain of the hyper-personalised network. It aggregates contextual information from devices, applications and the network itself, learns from past experiences and translates user intent into real-time adjustments across the radio access network, transport and core user plane. This includes dynamically managing slicing, policy, routing, edge resources and even sensing and computing capabilities.
From the device side, agentic user equipment plays an equally important role. Modern devices already have access to rich contextual information, including application behaviour, user activity and environmental signals. With embedded AI agents, devices can infer user intent locally and communicate it to the network in a way that respects privacy and security constraints. Rather than relying on deep packet inspection or static profiles, the network can act on explicit intent signals provided by the device itself.
This two-way negotiation between device and network addresses many of the challenges seen in 5G personalisation. It improves real-time feedback, reduces reliance on brittle heuristics and allows the network to anticipate issues rather than simply react to them. In principle, this could enable experiences such as seamless connectivity across transport modes, real-time optimisation for immersive applications, or guaranteed performance for embodied AI agents such as robots and autonomous systems.
However, the paper is clear that this vision is far from straightforward to implement. Standardisation remains a major open question. Existing network interfaces and APIs are not designed for rich, context-aware, agent-to-agent interactions. New mechanisms will be needed to support trust, discovery, semantic understanding and low-latency communication between AI agents across devices and networks. Without this, closed-loop hyper-personalisation will remain theoretical.
Trust is another critical challenge. Allowing AI agents to make autonomous decisions that affect network behaviour raises concerns around reliability, safety and unintended consequences. The paper highlights the need for strong assurance mechanisms, including runtime monitoring and fallback controls, to ensure that agentic systems behave within well-defined boundaries. This is particularly important in a telecoms environment where failures can have wide-reaching societal and economic impacts.
Energy consumption also cannot be ignored. Agentic AI systems require significant processing, both in devices and within the network. As operators already face rising energy costs, the efficiency of AI models, hardware choices and orchestration strategies will be crucial. Hyper-personalisation is unlikely to be viable if it dramatically increases the energy footprint of the network.
Despite these challenges, the direction of travel is clear. As AI becomes more deeply integrated into applications, devices and services, networks will need to evolve to support more dynamic, intent-driven interactions. 6G offers an opportunity to rethink the network architecture around this reality, placing agentic AI at its core rather than treating it as an add-on.
For those involved in 6G research, standardisation and training, the key takeaway is that hyper-personalisation is not just a feature but a systemic capability. It touches the radio, the core, devices, data models, interfaces and operational practices. Understanding this broader picture will be essential if 6G is to deliver experiences that feel genuinely different from what users have today, rather than simply faster versions of the same services.
Comments
Post a Comment