브레스저널 The Breath Journal

This article was translated automatically from the Korean original. Read the original in Korean

AI agent adoption has grown, but places actually running them stood at 11%

곽동현·Published 2026-10-14 12:00 KST
Financial Security Institute security indicators and research on agents verifying one another came out in the same week
AI agents that reached actual operation were reported at 11%
AI agents that reached actual operation were reported at 11% / ⓒ Breath Journal

The number of places bringing in AI agents has grown. But a report on October 8 said the share that has entered the actual operation stage is 11%. An AI agent is a program that breaks a task assigned by a person into several steps and handles it on its own. Many places run them on a trial basis, but only about one in ten actually hands them daily work.

The report did not make clear whether it counted companies or projects.

A day earlier, on October 7, the Financial Security Institute (금융보안원) released security indicators for AI agents. Security indicators are yardsticks that measure which risks were blocked, and to what extent, when an agent is attached to work. Agents access systems and act on behalf of people. In the financial sector, which handles money and personal information, deciding how far to extend permissions to an agent becomes a condition for moving to the operation stage, as much as the agent's performance.

On October 8, a cryptographic trust method that lets agents verify facts without having to trust one another was introduced in Cambridge. Cryptographic verification is a technique that uses mathematics to confirm that information handed over by the other party was not altered along the way. The method could be used where multiple agents exchange orders, payments and schedule coordination. An error in one place can spread down the line.

In recent workplace surveys, including those in the financial sector, senior executives and frontline employees differed in their views on AI use, confidence and trust. Even if the decision to bring in agents is made at the top, it is the front line that receives and checks the results every day.

The news that came out this week all dealt with control, verification and human trust.

If people have to reopen the money transfers an agent processed one by one, the work does not shrink. When tools such as the yardstick released by the Financial Security Institute take hold on the ground, and people only need to look at a few exceptions after machines finish verifying one another, the 11% figure will also begin to move.

Reporter Kwak Dong-hyun · Breath.Tech

Related articles

댓글