How AI Enables L4+ Autonomous Networks | Automation Demo at #MWC26

Summary of Technological Concepts / Product Features

• Autonomous Networks Fabric (unified layer): The demo introduces an autonomous networks fabric that provides a unified data and automation layer, combining:

  • Subscriber experience insights
  • Service assurance
  • Security operations

• Goal: L4/L5 autonomous operations: The system is positioned to help customers move from partial automation toward Level 4 and Level 5 fully autonomous network operations (“cross the chasm”).

• AI depends on data quality (data foundations): Speakers emphasize that AI is ineffective without good-quality data, so the fabric includes:

  • Curated data products
  • Data governance
  • AI lifecycle management, tracking:
    • Where data comes from
    • Who uses it
    • For what purpose
    • How decisions are made

• LLMs for network conversational interfaces: Large language models are used to reduce complexity and make network systems more conversational, enabling dialogue-like interaction with network functions.

• Generative AI to scale automation (human speed → machine speed): This enables:

  • Actions at machine speed
  • Removal of constraints tied to individual human capability
  • Swarms of agents (multi-agent automation)

• Safety controls for agent-based decisions: Agents must be prevented from making mistakes that could affect critical infrastructure. The fabric includes:

  • Guardrails and policies
  • Observability to keep agents within limits It’s described as a “glass box approach” to AI, emphasizing transparency and controllability.

• Closed-loop automation for security response: The demo showcases closed-loop automation, including:

  • Quickly determining root cause
  • Resolving/mitigating impact as fast as possible

• Demo scenario and geopolitical/security angle:

  • Scenario: A tennis tournament with VIPs and politicians, where a malicious attack aims to bring down the network and disrupt the event.
  • Real-world basis: References a SIM farm discovered near the UN building in New York (last year).
  • Outcome in the demo: The fabric identifies the issue as security-related (not a network performance problem), then triggers automation to mitigate.

Main Speaker / Source

• Andrew Barrel — colleague presenting the autonomous networks fabric demo