Current discussions of generative AI in creative and educational contexts often rely on tool-based or alien-intelligence metaphors, emphasizing control, and transactional prompting. This paper proposes an alternative framework—Human Relational Systems (HRS)—that reconceptualizes AI systems as human-derived intelligences operating across distinct representational and operational substrates, and argues that collaborative structure, rather than tool capability alone, shapes outcomes. Within the HRS framework, we distinguish three interacting forms of intelligence: Embodied Human Intelligence (EHI), representing human judgment, craft, and situated decision-making; Unembodied Human Intelligence (UHI), encompassing language-based, disembodied systems trained on human cultural artifacts; and Visual Human Intelligence (VHI), characterized by generative, spatial–temporal, and cinematic modes of representation. We propose that when these intelligences are deliberately integrated in a relational loop, they produce an amplified system-level outcome termed Joined Operational Yield (JOY), expressed as EHI + UHI + VHI = JOY. Drawing on practice-based examples from independent cinematic production workflows, we describe a triadic collaborative cycle where creative work moves iteratively across intelligences. In this cycle, EHI initiates intent through narrative and aesthetic goals; UHI supports ideation, scripting, and structural reasoning; and VHI generates visual sequences and cinematic interpretations. Outputs from VHI return to EHI for evaluation and refinement, completing the loop. Crucially, this process is not transactional or linear, but relational and iterative, relying on sustained context, role continuity, and shared goals rather than isolated prompts. The paper concludes by proposing HRS as a practical and ethical framework for hybrid human–AI collaboration, offering a transferable, practice-ready model for creative, educational, and research settings.