Choi Jae-sik, a chair professor at KAIST, stressed that companies should not stop at adopting artificial intelligence (AI) but must also change their performance evaluation and work structures. His assessment is that even if individual employees become more productive, this may not translate into overall company performance if the organization's operating methods remain unchanged.
In the opening lecture of the "2026 Federation of Korean Industries Executives' Jeju Summer Forum," held at the Lotte Hotel in Jeju on the 15th, Choi said, "Using AI raises individual productivity, but organizational productivity does not always rise along with it." He added, "In line with AI adoption, companies must change their goals, performance management, incentives, and even organizational culture."
He also said companies should not halt investment simply because their AI adoption success rate is low. He explained that if only full automation is defined as success, a significant number of projects would be classified as failures, but in actual industrial settings, sufficient results can be achieved even at intermediate stages such as monitoring, forecasting, and decision-making support.
Accordingly, Choi proposed that the level of AI autonomy in manufacturing should be broken down into stages such as monitoring, forecasting and diagnosis, decision-making support, conditional automatic execution, and full autonomy. "Companies need to create a roadmap for what stage each process is currently at and how far it will go in two years and in five years," he said. "The approach of trying to make all operations unmanned at once is not realistic."
He also assessed that Korea stands at an advantageous starting line in the "physical AI" market. He cited the fact that few countries possess both software and manufacturing foundations at the same time, and that Korea holds competitiveness in high-value-added manufacturing such as semiconductors, automobiles, and shipbuilding.
"The United States is strong in AI software but lacks a manufacturing base, and Japan's AI competitiveness is weak relative to its manufacturing capabilities," Choi said. "Since Korea possesses both software and manufacturing capabilities, it has strong potential to leap forward as a leading nation in physical AI." He noted that the government is pursuing a strategy to invest about 10 trillion won by 2030 for the AI transformation of manufacturing, and stressed that companies too should accept agentic AI and physical AI not as a temporary trend but as a survival strategy.
However, he diagnosed that Korea's AI infrastructure and service competitiveness still fall short. Because a significant portion of domestic users' generative AI queries are processed at overseas data centers, he said Korea must grow its domestic AI service ecosystem alongside securing graphics processing units (GPUs) and building data centers.
"Securing computing resources alone does not make a country an AI powerhouse," Choi stressed. "If we have efficient technology and competitive services, it is also possible to create AI services at domestic data centers and export them abroad."
