Presentation Information
[3Yin-A-58]A Proposal for Hybrid Automation to Improve Semiconductor Fab Operations Using AI Agents and a Configuration Management PlatformA Human-in-the-Loop–Based Incremental Automation Architecture
〇Makoto Kogoh1, Chiaki Morita1, Shiho Onodera1, Tadayuki Yoshida1, Tomoaki Mori1, Toshio Nakamura1, Asako Iwai1, Michiaki Tatsubori1 (1. IBM Japan)
Keywords:
Agent,Automation,Semiconductor,Smart Factory,Human-AI Collaboration
Operational tasks in semiconductor manufacturing fabs (fabs) require high reliability and quick response, yet still rely heavily on manual judgment and operation. Recently, approaches to leveraging AI agents to streamline fab operations have been explored. However, full automation by AI is currently challenging due to the high risk of handling unexpected equipment states and exceptions, necessitating cautious implementation. This paper proposes a hybrid configuration that combines configuration management and automation using Ansible, an IT operations automation platform for data centers, with AI agents, while retaining human judgment for final decisions. This approach is expected to significantly reduce operational burdens and enable the safe, gradual realization of automation.