Only One of Four Agentic AI Fields to Receive Additional Government Funding
Only one of the four agentic AI fields will receive additional government funding. Agentic AI refers to artificial intelligence that acts on its own judgment, and the Ministry of Science and ICT has designed the program around 'selection and concentration,' aiming to cut human intervention to 10 percent or less, with the project set to get underway in earnest this month.
Through the Institute for Information & Communications Technology Planning and Evaluation (IITP), the ministry released the evaluation results last month, in August: Medical Park will lead the medical ultrasound field, NC AI work innovation, SimSimi everyday companionship for citizens, and V-ENG physics-based simulation. Each lead company, together with its partners, will receive 4.5 billion won in first-round funding over 18 months. After a performance evaluation at the end of next year (late 2027), only one of the four projects will receive an additional 4 billion won.
The sticking point in applying a single yardstick to such divergent goals is the evaluation criteria. The government's request for proposals (RFP) shows that each field's objectives and expected effects are worlds apart. Everyday companionship aims to stabilize citizens' emotional well-being and boost social trust, while work innovation targets productivity gains in human resources, accounting, and finance regardless of company size. Physics-based simulation envisions AI agents taking charge of planning, execution, and analysis to automate design and verification processes in manufacturing, energy, and construction. The ultrasound field cites improved patient survival rates through easing the healthcare staffing shortage and shortening diagnostic wait times as its expected effect.
Disputes over evaluation criteria have a precedent. In 'Dokpamo,' the program to develop an independent foundation model, Motif Technologies — which topped a benchmark by global AI evaluation firm Artificial Analysis — failed to make the cut. It placed only fourth in the expert evaluation, the AI expert-user evaluation, and the general public evaluation alike, and Motif filed a formal objection with the Ministry of Science and ICT.
The competitive format has its merits, too. Rivalry among companies raises the polish of the final products and puts limited public funds to efficient use. A mid-to-long-term vision also underlies the design: rather than building multiple services, nurture one proven winner aimed at overseas markets. The remaining challenge is drawing up common criteria to compare and evaluate the four projects with different goals, and the first-round performance evaluation will proceed as scheduled at the end of next year.
