Agentic AI in Networking#

Ioannis Kaminaras, Kostas Gerakaris (Uni Systems)

• Why: Agentic AI — autonomous LLM agents that plan and execute multi-step network changes — is the emerging subfield where the PM's identity interests (agentic/workload identity) meet their networking interests. This is an early, practitioner-level look at the concept from a systems integrator (Uni Systems), useful for understanding what's being attempted in the field today (and what its authorization/trust-model problems are).
• What: An overview of agentic AI applied to networking operations (program text and deck metadata only — no abstract).
• How: Unknown; expect LLM-agent architectures (tool use, planning loops) applied to network operations tasks like configuration, troubleshooting, or assurance.
• Where applicable: Conceptual relevance to any network team; approach is vendor- and lab-stage, so treat claims of production maturity with skepticism. Notably, running agents that make network changes raises exactly the identity/authorization questions the PM works on — a natural cross-reference with the WebPKI talk.