This article proposes a novel method to solve the manufacturing scheduling problems in multi-agent system (MAS) for dynamic environment. The study focuses on the machine selection work for jobs and proposes a remaining time prediction method, which will help a job choose machines based on estimated finishing time. In this way, we can keep the job finishing just-in-time, which aims to support a customer-oriented lead time policy, and can provide better manufacturing performance and increase customer service level. This article provides experimental results compared with another approach in previous research, through which we can find that in some case if we put the Just-in-time philosophy into consideration, the proposed method will deliver competitive performance.