AWS 联合 Strand Partners 发布《Unlocking Korea's AI Potential 2026》调查：韩国企业 AI 采用率升至 58%、但仅 18% 推出 AI 新产品，physical AI 仍处早期

AI adoption is spreading so quickly in South Korea that more than one company begins using the technology every minute. Yet only 18% of firms have launched a new product or service built on AI, indicating that technology adoption is translating into broader business innovation at a relatively slow pace. Analysts say that for Korea — with its solid base in manufacturing, semiconductors and robotics — to seize an early lead in the still-nascent physical AI market, it must simultaneously raise its AI strategy, data and infrastructure, and workforce capabilities.
Amazon Web Services presented these findings on the 11th at its AWS Korea office in Yeoksam-dong, Gangnam District, Seoul, in a report titled "Unlocking Korea's AI Potential 2026."
The report was based on a survey by Strand Partners, a global policy and economic consultancy, of 1,000 business leaders — including founders, chief executives and C-level executives at Korean companies — and 1,000 members of the general public.
The survey found AI adoption among Korean companies rose 10 percentage points to 58% this year from 48% a year earlier. Nick Bonstow, a director at Strand Partners, said about 624,000 Korean companies newly adopted AI over the past year, which works out to more than one company every minute. He added that 81% of firms that adopted AI reported gains in productivity, saving an average of 14 hours a week.
The sophistication of AI use has also risen a notch. The share of companies deploying advanced AI more than doubled to 26% this year from 11% a year earlier. Meanwhile, the share of firms using AI mainly for basic, incremental improvements to existing work fell to 54% from 70% over the same period, showing that corporate AI use is moving beyond simple task assistance toward more advanced applications.
Still, AI adoption has been slower to spread into company-wide business innovation. Among firms that have adopted AI, only 18% have launched an AI-based product or service. Just 27% have a formal AI strategy, and 24% have a consistent framework for measuring results. Notably, 47% said they have no reliable way to measure return on investment from AI. While the number of companies adopting AI is rising fast, the systems for measuring returns and scaling successful cases across an organization are not yet sufficiently in place.
Gaps also emerged by company size. Startups had the highest rate of advanced AI adoption at 35%, followed by small and medium-sized enterprises at 25%. Large companies stood at just 9%. At the same time, 84% of firms expect their AI use to increase over the next 12 months, and 72% expect AI to transform their industry. That suggests willingness to use AI is high, but the pace of actual transformation across organizations and businesses varies widely by company.
In the next generation of AI markets, Korea's manufacturing base is expected to open new opportunities. AWS singled out physical AI as an opportunity particularly suited to Korea.
Bonstow said 6% of Korean companies have fully deployed physical AI, while 22% are running proofs of concept or pilots and 39% plan to adopt it in the future. Because the market is still in its early stages, he said, Korean companies with competitive strengths in manufacturing, robotics, semiconductors and advanced technology have ample room to establish an early lead. The areas companies cited as the main opportunities in physical AI were also concentrated in fields closely tied to manufacturing: autonomous robots (54%), predictive maintenance (41%), human-machine collaboration (39%), warehouse and logistics automation (38%) and intelligent supply chains (34%).
Agentic AI also remains an early-stage market, though results are emerging at companies that have adopted it. Some 37% of firms said they were familiar with agentic AI, and among them 8% have fully deployed it while 29% are running proofs of concept or pilots. Of the companies that adopted agentic AI, 54% said decision-making and execution had grown faster, 51% cited gains in operational efficiency and productivity, and 38% reported improved operational scalability.
Shinhan Financial Group, working with the AWS Generative AI Innovation Center, built an agentic AI system based on Amazon Bedrock AgentCore and saw tangible results, including a sharp improvement in the accuracy of classifying customer intent. The system, which sorts customer intent across 12 domains, 12 actions and more than 28 tags, raised intent classification accuracy to between 92.5% and 97.9%. Accuracy in linking actions reached 100%, and inference costs were cut to one-eighteenth of previous levels. Overall response latency was also reduced to under five seconds.
Oh Soon-young, a senior solutions architect at AWS, said agentic AI would expand beyond supporting corporate work to act as an economic agent in its own right. "An era is coming in which the entity that sets product prices, builds supply chains, issues refunds to customers and purchases the data it needs could be an AI agent rather than a person," the architect said. "Calling agents simply tools is now too narrow a description."
The question is how prepared companies are to embrace next-generation AI. Only 24% of all firms said they were sufficiently or very well prepared for next-generation AI such as agentic AI and physical AI. Among startups the figure was relatively high at 43%. Companies cited a lack of AI and digital capabilities (46%), technology and data barriers (38%), difficulty securing enough talent (33%) and limited access to tools and infrastructure (28%) as the main obstacles.
There was also a strong sense that AI competitiveness will hinge on physical infrastructure such as electricity and data centers, not just models and software. Some 74% of companies said a stable power supply is important to long-term AI investment. Access to renewable and carbon-free energy was cited as important by 63%, and long-term cost predictability by 60%. Asked what is needed to make Korea more attractive for AI and digital infrastructure investment, companies pointed to greater predictability in securing power (46%), expanded grid and transmission capacity (42%), greater predictability in infrastructure construction timelines (38%) and shorter permitting procedures (35%).
AWS assessed that the next task for Korean companies is to move beyond AI adoption itself and turn it into repeatable business innovation. Bonstow stressed that companies must build clear AI strategies and performance measurement frameworks, strengthen their cloud and data foundations, and expand infrastructure investment in AI computing environments and a stable power grid. Preparing the workforce, including expanding AI and digital skills and providing continuous retraining, is also a core task, he said.