From Task Assistant to Goal-Sharing Colleague: The Agentic AI Era Is Coming [EastPoint: Seoul 2026]

Kim Ho-jin, CEO of Hashed Open Finance and ShardLab

“I'm not a robot.”

That line appears several times a day as websites try to determine whether a user is a person or a bot, an automated software program. For more than 30 years, the web has been built for humans. Developers focused on improving the user experience for people. Security systems distinguished humans from automated programs. Financial authentication verified that the person making a transaction was the rightful owner of the assets. Search, shopping and payments all rested on one assumption: a human was acting directly.

Before long, we may start seeing the opposite message more often: “I'm not a human.”

The force challenging that assumption is the AI agent. It is software that can search on its own, make reservations, carry out tasks and collaborate with other AI systems. This new paradigm, in which AI actively pursues goals, is known as agentic AI.

The key phrase is active execution. Generative AI has so far served as a support tool that answers questions. AI agents understand a goal, make a plan, move across multiple systems and complete the job. A single instruction such as “Prepare my business trip” can prompt an agent to find flights, book a hotel, coordinate a schedule and draft the necessary emails. The internet is gaining a new actor: AI that carries out goals rather than simply answering questions.

This is not a distant future scenario. In Microsoft's Work Trend Index report last year, based on a survey of about 30,000 workers in 31 countries, 81% of business leaders said they planned to integrate agents into their AI strategy within 12 to 18 months. IDC projects the number of AI agents operated by companies worldwide will surpass 1 billion by 2029.

I see this shift not as a simple technology trend, but as the starting point of a broader transition in which society, the economy and infrastructure are redesigned.

Work will change first. Microsoft says that as AI agents take on more execution, people's roles will shift toward setting goals, making judgments and taking responsibility for outcomes. The company calls those workers “agent bosses.” In effect, people who once carried out tasks themselves will become supervisors who manage and review AI's work. That also means the unit of labor will be reorganized around goals rather than individual tasks.

A more interesting change is unfolding in the market economy. Until now, companies have devoted their efforts to persuading people. They refined advertising and marketing. They polished interfaces and payment buttons to make them easier for humans to use.

But once agents handle search, comparison, purchasing and payment, companies will find non-human customers on their client lists. Flashy banner ads mean nothing to an agent. What matters is whether product information, prices, inventory and return policies are organized in a machine-readable format, and whether payment and settlement can be processed automatically. Information gaps will also narrow sharply. A world in which everyone's agent compares and buys products with near-complete information moves much closer to the perfectly competitive market described in introductory economics textbooks, where efficiency is maximized. A strong consumer brand alone will no longer be enough. Data and interfaces that can win selection by agents will define a new set of competitive rules.

In global fintech and blockchain markets, experiments are already underway in which agents make payments with stablecoins and complete settlement on their own. Those tests are only a preview of a much larger shift. A change in who acts requires new infrastructure. If AI is to act on behalf of humans, it must be able to prove whom it represents, verify what authority has been delegated to it, make payments on its own and validate transaction results. That is why technologies such as decentralized digital identity and permission management, stablecoins, programmable payments and agent-to-agent communication, or A2A, are drawing attention together. In the agent era, a new digital infrastructure must be built so people can confidently entrust work to AI.

This presents a rare opportunity for South Korea. Few countries are as well positioned to build a trust-based transaction environment for agents, with a financial system built on real-name accounts for the entire population, the government's push to become one of the world's top three AI powers, full-stack sovereign AI investment spanning semiconductors to domestic models, one of the world's highest adoption rates for simple payments, and ongoing discussions in the National Assembly and the government on institutionalizing won-denominated stablecoins. What is needed now is not another AI service. It is a foundation that ties those conditions into a single infrastructure and positions the country to lead the standards of a new era.

That thinking is already spreading across industry. Hashed, Bloomingbit and the Korea Economic Daily chose “The Agentic Era” as the main theme of this year's global conference, EastPoint: Seoul 2026, in the same spirit.

Before long, websites may ask visiting agents a new question: “Whom are you here on behalf of?” The country that first builds infrastructure capable of answering that question will shape the internet order of the next 30 years. The biggest change in how work is done since the birth of the internet has already begun.

Kim Ho-jin, CEO of Hashed Open Finance and ShardLab