Naver D2SF said on May 19 it made a new investment in Clonlabs, a developer of a “user model” aimed at solving fatigue issues for artificial intelligence (AI) agents.
The investment was made in swift fashion just three months after Naver D2SF discovered the team through a campus tech startup contest. It reflects a high assessment of Clonlabs’ ability to solve the newly emerging “human-agent bottleneck” in the AI era.
As use of AI agents grows, work fatigue is increasing because users must repeatedly issue commands and provide feedback. Clonlabs is developing a user model that predicts context and makes decisions on its own without user involvement. In the structure, an AI that has learned a user’s computer-use patterns detects when the agent stops and carries out the next action instead.
To improve model accuracy, Clonlabs applied three layers: “recording” that logs computer-use patterns, “memory” that analyses context and preferences, and “prediction” that forecasts and executes the next action. It also sought stability in automation by asking users for confirmation only when prediction confidence is low. As its first target, it introduced “Clone Desktop” and “Clone Plugin” aimed at “AI builders,” who use AI agents the most.
The team’s research capabilities are also solid. Made up of Seoul National University undergraduate alumni, Clonlabs has seven related papers through collaborations with Stanford University and Carnegie Mellon University, among others. Its strengths include experience in in-depth research on core technologies such as AI that directly controls computers (CUA, Computer-Use Agent) and privacy-preserving memory architectures.
Yang Sang-hwan (양상환), head of Naver D2SF, said new bottlenecks and opportunity markets that did not exist before are emerging as AI technology advances. He said Clonlabs is a team that identified the problem that bottlenecks between people and agents limit productivity and is solving it with focus.