The government is officially launching efforts to secure core technologies for 'Agentic AI'—AI that autonomously sets tasks and makes judgments and actions without human intervention.
The Ministry of Science and ICT and the Information & Communications Technology Promotion Institute (IITP) announced on the 30th that they held a kickoff meeting for the "Development of Real-World Active Agentic AI Technology" project at the Seoul AI·SW Maestro Center. This project is a key mission under the government's recently announced "Agentic AI Initiative," aimed at building a domestic Agentic AI ecosystem and strengthening technological competitiveness.
While Korean companies have previously focused on simple AI assistants and other digital agent services, this R&D project aims to combine digital spaces with the physical real world, reducing human intervention rates to 10% or less. It will promote the commercialization of advanced Agentic AI services that autonomously plan and execute complex tasks in the real world.
The government plans to invest a total of 18 billion won over one and a half years from this year through the end of 2027. Four lead institutional investors—NC AI, Medical Park, Simsimi, and VENG—are participating, along with a total of 26 industry-academia-research institutions. After a final evaluation at the end of 2027, one outstanding project will be selected for an additional year of support to complete a prototype at the global top level.
Specifically, projects in four areas will be carried out: △Agentic AI Breast Ultrasound (autonomously performing tasks from image acquisition and analysis to report writing at the level of skilled human examiners), △Agentic AI Digital Work Colleagues (understanding and learning distributed work processes such as ERP, email, and calendars to autonomously solve problems), △Daily Companion and Empathetic Colleague (providing empathetic interventions and practical life support tailored to user situations), and △Agentic AI Simulator (automatically performing physical interpretation through multi-agent collaboration based on the MCP standard).
Participating institutions plan to develop prototypes at the "TRL 7" level, which can demonstrate tangible results beyond the research stage and are ready for demonstration and commercialization. To achieve this, they will focus on overcoming key technical limitations including step-by-step planning and revision, long-term memory and context management, multi-agent collaboration, and safety and alignment.
Lee Jin-soo, Deputy Minister of AI Policy at the Ministry of Science and ICT, stated, "In the era of Agentic AI, securing real-world judgment and execution technologies is as essential as acquiring superior models," promising active policy support.
Hong Jin-bae, Director of IITP, also announced, "As the Agentic AI market and technology develop rapidly, we will apply a moving target management approach to allow research teams to focus on technology development."
in TECH
