LLM Architectures for Agentic Network Operations and AIOps
A recent paper on arXiv (2605.12729) reviews the implementation of large language models (LLMs) in network operations (NetOps) and AI for IT operations (AIOps). The authors categorize existing research based on a framework that includes autonomy levels, tool capabilities, evidence documentation, and assurance agreements. These agreements detail the observations, proposals, and actions permitted for an agent, along with necessary checks prior to execution. Applications discussed include telemetry query suggestions, incident analysis, root-cause identification, configuration generation, and basic self-healing. The study highlights that operational choices have immediate consequences, underscoring the importance of safety and assurance. This paper can be found on arXiv under the cross type.
Key facts
- Paper arXiv:2605.12729v1 focuses on LLMs for NetOps and AIOps.
- Covers incident investigation, root-cause analysis, configuration synthesis, and self-healing.
- Introduces hierarchy of autonomy, tool scope, evidence traces, and assurance contracts.
- Assurance contracts define what an agent may observe, propose, and execute.
- Agent workflows include permissions, policies, checks, and rollback options.
- Operational decisions have instant impacts, requiring safety measures.
- Literature organized around telemetry query recommendation and diagnostics.
- Published as a cross-type announcement on arXiv.
Entities
Institutions
- arXiv