Engineers Move Beyond Writing Code to Designing AI Workflows

According to AI TIMES, a column by Samsung SDS manager Jeong Cheon-su, introduced on September 9, points out that in an era when AI writes code, engineers' core role is shifting from writing code directly to designing how AI works. He said that even though related terminology has changed rapidly in less than half a year since harness engineering drew attention, the essence lies in establishing role division between humans and AI and a validation structure. The column follows the flow of explanations expanding from prompt to context, harness, agentic, and loop engineering, and concludes that the focus should be on direction rather than names. It argues that while AI can handle planning, execution, and iterative improvement, humans must set goals, determine acceptable boundaries, and retain final responsibility. In the end, the piece says engineers are being redefined as those who separate tasks to assign to AI from tasks requiring direct judgment, and who design the necessary information, procedures, and checkpoints. It emphasizes that the future competitive edge will come not from increasing AI's autonomy itself, but from how well human judgment and AI's execution capability are combined and stably applied to real work.