"Now, not only people but also software with a purpose and a budget will become customers."

Oh Sun-young, principal solutions architect at Amazon Web Services (AWS), said this at a press briefing for the release of the report "Unleashing Korea's AI potential in 2026," held on the 11th at the AWS Korea office in Centerfield in Yeoksam-dong, Gangnam-gu, Seoul. Oh said 51% of current web traffic comes from bots, and that an "agent economy," in which AI agents rise as the subject of transactions, is becoming a reality.

Oh Sun-young, senior solutions architect at AWS, speaks at a press briefing for the release of the report 2026 AI Potential of Korea at the AWS Korea office in Centerfield, Yeoksam-dong, Gangnam-gu, Seoul, on the 11th. /Courtesy of AWS

◇ A market opens where AI buys and sells data

AWS is rolling out capabilities that let AI agents participate directly in transactions. Bedrock Agentcore Payments, which AWS officially launched last month, allows agents to search for and pay for paid APIs (application programming interfaces), MCP (Model Context Protocol), and content on their own. Within allowed tools, amounts, and conditions, agents use Coinbase and Stripe to pay autonomously. Suppliers such as digital content owners and publishers can charge AI bots directly through AWS's web application firewall (WAF).

Oh said, "The main actors of web activity are shifting from people to machines," adding, "A single piece of data or an API call can become a digital product purchased by a machine." Oh went on, "As agents transact, the unit of price will also change," noting, "Right now we only talk about tokens, but in a market where machines purchase on their own, the basis of expense will shift from tokens to 'completed outcomes.'" In other words, the charging unit will be not how much AI was used, but "how much of the user's task was actually accomplished."

Oh said the software development kit (SDK) for AWS's agent management tool Bedrock Agentcore has been downloaded about 2 million times just five months after being made public, adding, "The moment software adjusts goals, resources, and transactions on its own beyond answers, AI becomes a new economic actor, not a tool." Oh also projected that the share of agents integrated into corporate applications (apps) will grow from under 5% at the end of 2025 to 40% by the end of this year.

◇ AI adoption in Korea has increased, but only 27% of corporations have a strategy

The agent economy vision Oh outlined was based on the judgment that AI adoption by Korean corporations has matured accordingly. The report "Unleashing Korea's AI potential in 2026" formed the basis for discussion of the agent economy. According to the report, in which research firm Strand Partners, commissioned by AWS, surveyed 1,000 domestic corporate leaders and 1,000 members of the general public, AI adoption among Korean corporations rose to 58% this year from 48% a year ago. Over the past year, about 624,000 corporations newly adopted AI.

The expansion of AI adoption has created real value. Among corporations that adopted AI, 81% reported productivity gains and 58% reported sales growth. They saved an average of 14 hours per week with AI. The share of "basic-stage corporations" that use AI only for core tasks fell from 70% to 54%, while the share of "advanced-stage corporations" that combine multiple models or deploy autonomous AI more than doubled from 11% to 26%. However, only 27% of corporations had formal, comprehensive AI strategies, and just 24% had consistent criteria to measure AI success.

Ham Ki-ho, head of AWS Korea, said, "Korean corporations have built significant momentum in AI, and many are already delivering tangible results in productivity and growth," adding, "The question now is not whether to adopt AI, but how to scale it across the business." Ham noted, "Organizations with a solid foundation across data, capabilities, governance, and technology will be in the strongest position."

※ This article has been translated by AI. Share your feedback here.