News

From Reactive Assistants to Agentic AI

We are moving beyond simple voice commands and reactive apps toward Agentic AI, intelligent systems that act autonomously on our behalf. These agents can manage schedules, monitor health, or navigate complex environments without constant input. They rely on the cloud for model updates, data aggregation, and cross-device synchronization. But when speed is essential, intelligence runs locally on the device, leveraging edge computing for real-time responsiveness.

A major challenge in deploying AI at scale is protecting user privacy. This is where Federated Learning comes in. Instead of collecting raw user data, federated learning allows AI models to be trained across thousands of devices using only encrypted model updates. The raw data never leaves the device, ensuring privacy and compliance with regulations like GDPR. This technique is now widely used in mobile keyboards, health apps, and personalized recommendations [2]. The NOUS project leverages this approach to build privacy-preserving AI systems that learn from real-world usage without compromising user trust.

Enabling Smarter Agent Collaboration

As intelligent agents become more common, they need to communicate, not just with users, but with each other. This is the promise of Agent-to-Agent (A2A) communication, where autonomous systems collaborate, negotiate, and coordinate actions without centralized control. Imagine autonomous vehicles warning each other about road hazards, or smart home agents adjusting schedules based on shared context. A2A enables decentralized, resilient, and scalable coordination in dynamic environments [3].

But for agents to truly understand one another, they need more than just data. They need context. The Model Context Protocol (MCP) is designed to solve this. It allows AI agents to exchange not only alerts or commands but also situational awareness, intent, confidence levels, and constraints. This shared context enables more meaningful, adaptive interactions, a key requirement for smart cities, robotics, and immersive digital environments. While MCP is still emerging, it builds on the growing field of context-aware computing, where AI systems interpret and respond to their environment in real time [4].

The Critical Role of 5G Connectivity

None of this would be possible without ultra-fast, low-latency connectivity. This is where 5G networks come into play. With speeds up to 10 Gbps and latency as low as 1 millisecond, 5G enables seamless communication between edge devices and the cloud. It supports massive device density, making it ideal for IoT and mobile edge AI. Crucially, 5G provides Ultra-Reliable Low-Latency Communication (URLLC), which is essential for safety-critical applications like autonomous driving and remote surgery [5]. Without 5G, real-time A2A messaging and dynamic edge inference would not be feasible at scale.

One compelling use case being explored by the NOUS project involves connected vehicles using camera data for accident prevention. Onboard AI processes video in real time at the edge to detect pedestrians, traffic signals, and obstacles. When a hazard is identified, the vehicle can instantly alert nearby cars via A2A communication, using MCP to share not just the alert but also the confidence level and environmental conditions. Model improvements are shared securely through federated learning, and 5G ensures these messages are delivered in time to prevent collisions. This integration of edge, cloud, AI, and high-speed connectivity showcases the future of intelligent mobile systems.

Conclusion

Edge and cloud computing are not competitors but collaborators in enabling Agentic AI for mobile networks. The optimal balance depends on:

• Latency requirements (edge for real-time, cloud for batch processing).
• Data sensitivity (edge for privacy, cloud for large-scale analytics).
• Cost & scalability (cloud for elasticity, edge for localized efficiency).

As 5G/6G, AI and distributed computing evolve, the line between edge and cloud will blur, creating an intelligent, adaptive ecosystem where Agentic AI thrives.