AI Agents Expanding in Network Operations
As AI agent adoption in network operations expands, organizations face the challenge of ensuring that agents operate reliably and predictably. How to balance the convenience of autonomous AI decision-making and behavior with the 'deterministic operations' required in infrastructure management has emerged as a common industry challenge.
The use of AI agents is expanding in network operations. However, while many organizations seek to deploy AI more widely, the difficulty of ensuring that agents operate in a "reliable and predictable manner" has emerged as a common industry challenge.
An AI agent refers to an AI system that makes decisions and takes actions autonomously without waiting for human instructions. In the context of network management, tasks such as fault detection, configuration changes, and traffic optimization were traditionally handled by human engineers. AI agents can automate such work, attracting the interest of companies expecting cost reduction in operations and faster response times.
However, network operations is a domain where errors can directly lead to service disruptions or information leaks. For this reason, how to balance the "autonomy" of AI agents in making flexible judgments based on circumstances with "determinism" (deterministic behavior) that operates as administrators expect has become the primary focus in deployment. Determinism refers to the property of always returning the same result for identical inputs, which is especially important in infrastructure management.
The tradeoff that increased AI usage makes agent decision-making harder to predict is not unique to the network field. It is also a common challenge for generative AI in general. However, in network operations, the impact range of failures is wide and recovery can take time, so the requirement for "predictable behavior" is positioned as stricter compared to other fields.
When organizations incorporate AI agents into network operations, design is required that predefined agent action scopes and clarifies the range of operations requiring human approval. Without establishing a governance framework that distinguishes between tasks to delegate to agents and tasks where humans make final decisions, it is difficult to enjoy the benefits of automation while maintaining safety.
The deployment of AI agents in network operations is a development that could serve as an opportunity to fundamentally reconsider the approach to infrastructure management. The discussion around autonomy and determinism is expected to deepen with technological maturity, and how agent design philosophy and governance standards are developed becomes a point of future attention.
This article is an original work independently written and edited by the AI issue editorial team based on factual reporting. © AI issue. Unauthorized reproduction, redistribution, or use for AI training is prohibited.