Working Paper · September 2026
The Clock, Not the Culture — Predicting Workplace Behavior from the Individual to the Multinational in the Age of AI Colleagues
“It is not the culture; it is the decision latency.”
한 사람의 직장 내 행동을 — 약하고 조건부이지만 측정 가능하게 — 예측할 수 있다면, 팀의 행동은? AI 동료가 포함된 팀은? 기업은? 다국적 기업은? 이 논문은 문헌이 가장 빈약한 수준에서 원자료를 제시한다. 조직이 사람에게 쓰는 도구 — DISC 형식 검사, 빅 파이브 설문, 반복 게임 배터리 — 를 실제 직무가 구성하는 방식대로 구성된 AI 동료에게 적용한 9,772건의 측정이다. 결론은 세 가지다. 조직이 자문가에게 — 인간이든 인공지능이든 — 부여하는 역할은 국가 문화가 아니라 의사결정 지연 시간을 따른다. 문서화된 힘은 예측하지만 문서화되지 않은 힘은 예측하지 못한다. 직무는 AI 동료의 스타일을 쓰고, 모델은 그 행동을 지닌다.
If one person’s behavior at work can be predicted, weakly, conditionally, but measurably, can a team’s? A team that includes an AI colleague? A company’s? A multinational’s? Five literatures each hold a piece of the answer and none holds the whole. This paper asks the question once, across all five levels, and finds that prediction does not fail as you climb; it changes its translation rule at every boundary.
The paper contributes original data at the level where the literature is thinnest: 9,772 measurements administering the instruments organisations use on people, a DISC-format inventory, a Big Five questionnaire and a battery of repeated-play games, to AI colleagues configured the way real functions configure them: six corporate roles, working documents included, across four model versions from two providers.
The role an organisation assigns to its advisors, human or artificial, follows its decision latency, not its national culture. Written strength predicts; unwritten strength does not. The function writes the AI colleague’s style; the model carries its behavior.
Characterise the situation before reading anyone’s profile. Design the team’s first round instead of predicting it. Contract the advisor’s role rather than letting the clock assign it. And before an AI colleague joins a team, profile it like a hire: per role, per model version, with the behavioral battery and not just the questionnaire, then re-profile on every update.