arXiv:2611.25269v1 Announce Type: new Abstract: Generative AI is becoming embedded in knowledge work as an intelligent collaborator to assist in completing tasks. However, this readily available assistance risks displacing key interactions through which colleague relationships are built and sustained at work. This paper leverages the concept of help-seeking behaviour to investigate whether and how AI usage shapes colleague relationships. We conducted an online experimental vignette study where 248 knowledge workers evaluated scenarios that systematically varied whether and how AI was used across different work tasks. Using linear mixed-effects models, we found that AI-enabled work reduced help-seeking intentions compared to non-AI work across tasks, and especially for brainstorming. Visible AI usage was associated with greater help-seeking intentions towards colleagues, and individual contributors' colleague relationships were more vulnerable than managers. Our findings identify opportunities for workplace AI tools, practices, and policies to preserve colleague relationships in increasingly AI-integrated workplaces.