This study explores help-seeking supported by Generative Artificial Intelligence (GenAI), peers and the teacher during problem-solving sessions. The research aims to collect and analyze data from various sessions in a vocational training setting to compare which are the most effective doubt-resolution strategies for tasks related to software programming. By involving students in the analysis, the study compares their perceptions of the efficacy of responses from GenAI tools, peers, and teachers against their initial expectations. Results indicate that students unexpectedly rate the GenAI tool’s responses more favorably than those from their peers and the teacher. This paper discusses these findings, informing the opportunities for effective integration of GenAI tools in educational settings depending on the context. The comparative analysis seeks to illuminate the complementary roles of GenAI tools, peers, and teachers in the design of human-machine co-orchestration strategies involving help-seeking in problem-solving classrooms.