What We Can Learn from AI Coding Agents — Raising a Glass-Box AI Agent for NetOps – Net One Triple-T
In 2026, vendors rapidly released AI-agent-driven network operations tools. Yet relying on AI in operations demands clarity: what logic and which knowledge led to each recommendation? Our customers in Japan want to know how to detect and correct an agent’s wrong judgment, how to teach it, and how to roll a mis-trained agent back to a previous state. We propose applying practices established in AI coding agents to network operations, illustrated with a demo.
Masahiro Yoshida is an engineer at NetOne Systems with over 10 years of experience in the IT industry. Starting out as an application developer and now working across both engineering and research and development, he leads research into autonomous network operations, bringing a software engineering perspective to network infrastructure. His current work focuses on building prototypes and validating them against real-world operational scenarios.
Suguru Kurihara is an engineer at Net One Systems with over 20 years of experience in the IT industry. Having built his career as an open-systems application developer — primarily in Java-based web application development — he joined Net One Systems in 2018. Since then, he has been involved in the development and operation of in-house applications, as well as the research and validation of network automation and autonomous networking services, bringing a deep software engineering perspective to the evolving field of network automation and autonomy.
