South Korea Plans Talent Policy for Agentic AI-Driven Software Era

South Korea’s Ministry of Science and ICT says agentic AI could dramatically shorten software development timelines, with projects that once took years potentially reduced to about 40 days. The ministry described this as a sign of a fundamental shift in how software is built as AI becomes more autonomous in its work.

To prepare for that shift, the ministry has held six expert meetings in the past month under the theme “Agentic AI era, SW industry and talent development responses,” drawing roughly 70 professionals from industry, academia, and research institutes. The aim is to translate discussion into policy guidance for the AI-enabled software economy.

Industry observers note that development teams are already changing, with fewer developers directly reporting to a single product manager as AI-assisted coding alters team structures and workflows. Experts say talent development must keep pace with these changes.

Participants stressed the need for new talent profiles suited to an agentic AI era, emphasizing skills to define problems, design systems, and verify outcomes, all while leveraging AI as a core tool. The shift is seen as moving the emphasis from simple coding toward high-level system thinking.

Song Ho-cheol, CEO of Duzon Bizon, told the gathering that AI now handles much coding work, so developers are spending more time designing and defining how AI operates, signaling a fundamental change in the developer’s role.

Education leaders at the meeting argued for a full curricular overhaul—moving beyond coding drills to curricula that stress design, verification, and system architecture. Universities should prioritize core software fundamentals and architecture to teach the essence of software development in an AI-powered environment.

The panel highlighted two talent tracks: advanced engineers with deep fundamentals and “convergence” or fusion talents who combine domain knowledge with AI skills. Kakao Enterprise PM Wang Ji-yeong urged universities to emphasize cross-department courses and project-based experiences that let students tackle real-world problems rather than merely forming new departments.

On the infrastructure side, the discussion touched on GPU clusters for AI education. KAIST professor Seong Min-hyuk noted that Korea lacks school-level GPU clusters for AI training, suggesting options such as repurposing older corporate GPUs and separating research and education, or building shared clusters across multiple universities to spread costs.

The government plans to draft a comprehensive talent development policy tailored to the agentic AI era, aiming to align higher education, industry needs, and research ecosystems. The remarks were delivered by Ryu Jae-myeong, deputy minister of MSIT, at a National Science and Technology Advisory Council meeting in the Kyobo Building, Gwanghwamun, Seoul, on March 13, 2026.