AI-based online testing platform company Grepp said on Wednesday it has applied an AI agent based on a large language model (LLM) to its online exam proctoring solution Monito.

The company said the newly added AI agent panel in Monito is placed on the right side of the proctoring screen and delivers analysis results in real time. While the existing system only alerted when a physical event occurred, such as a hand moving out of the screen, the new AI agent analyses behaviour patterns and reports that a test taker repeatedly looked away during the exam and stared in a certain direction, or that proxy test-taking and referencing a secondary monitor are suspected.

The company said internal tests showed that reviews based on AI-generated narrative reports cut post-review time by more than about 30 percent, compared with a method that conducted a full review of the entire recorded video.

Grepp Chief Executive Lim Sung-soo (임성수) said, "The most difficult part of exam proctoring is determining whether a certain behaviour is actually cheating." He said, "If AI explains the behavioural context and presents objective grounds, proctors can make final decisions faster and more accurately. This update will address that on-the-ground difficulty."

Grepp plans to further advance its proctoring operating system in the first half of 2026. It plans to introduce a "dual feed" structure that lets customers choose based on exam type, applying a live feed method in which human proctors and AI collaborate for exams requiring high reliability, and an AI feed method in which an AI proctor handles cases requiring large-scale post-review.