Mental healthcare systems face increasing pressure from rising demand, fragmented services, limited resources, and models still organized around episodic clinical encounters. This article examines whether artificial intelligence (AI) can be conceptualized not pri-marily as a digital therapy, chatbot, or substitute for professionals, but as a supervised infrastructural layer supporting continuity of care within distributed mental healthcare communities. The study adopts a conceptual, systems-oriented design informed by structured evidence mapping of recent literature on AI in mental health. The evidence map was based on a structured PubMed search from January 2022 to March 2026, from which 33 analytically relevant studies were selected and classified according to clinical function, study type, clinical context, level of clinical integration, and systemic relevance. The analysis applies a systems lens focused on actors, boundaries, data and information flows, decision and escalation points, feedback loops, and governance conditions. The mapped literature is reorganized around five emerging clinical functions: longitudinal observation, conversational orientation and support, functional inference, clinical decision support, and assisted interventions. Read systemically, these functions may help address discontinuities between visits, support interpretation of patient trajectories, and connect everyday environments, professional judgment, and care responses. However, the field remains clinically immature, with limited integration into real workflows and persistent risks related to accountability, surveillance, bias, and overdependence on poorly vali-dated systems. The article argues that AI can contribute to trajectory-based mental healthcare only when embedded in human-led workflows, escalation protocols, com-munity-based care architectures, and institutional governance.