Design and Evaluation of a Multi-Agent Conversational AI for Students’ Support-Seeking during Workplace Onboarding

Students often hesitate to seek support during workplace onboarding. While AI systems are
increasingly used for advice and brainstorming, less is known about how conversational inter
actions can reduce support-seeking inhibition in workplace contexts. This study investigates
the challenges and needs that contribute to inhibited support-seeking among newly hired stu
dents during workplace onboarding and examines how conversational interactions with LLM
powered agents can be designed to support support-seeking in this context. A multi-agent con
versational prototype was evaluated in a within-subjects study with 30 participants. The findings
show that conversational interactions can mediate students’ support-seeking by preparing them
for human interaction in both task-related and interpersonal workplace situations. This occurs
when the system structures uncertainty into manageable steps, regulates hesitation through au
thentic empathy, provides safe rehearsal spaces, and supports reflective decision-making. The
study contributes a problem-based conversational interaction model and a design framework for
student–AI collaboration that supports support-seeking during workplace onboarding.