With the advancement of digital technology and Industry 4.0, artificial intelligence (AI) is gradually embedded in human resource management and has become an important digital foundation to support the sustainable transformation of enterprises. However, the research in the manufacturing context, particularly through the challenge perspective at different levels, remains fragmented. This work represents a systematic review of 347 articles from Scopus and Web of Science from 2000 to 2025 and employs a dual-method analysis strategy embracing metrics and in-depth coding on 100 core publications. Excel, Bibliometrix, CiteSpace, Latent Dirichlet Allocation (LDA), and VOSviewer were utilized for quantitative analysis, while open–axial–selective coding of the Grounded theory approach was applied to generate qualitative results. The findings revealed six key challenges in integrating AI-HRM within manufacturing and six approaches to solve the identified issues. The Challenge–Approach Matching Matrix was constructed, illustrating the suitability of different pathways for addressing specific challenges. Analysis of thematic evolution in AI-HRM research resulted in the identification of three distinctive phases and demonstrated a consistent shift from technology-centric approaches towards human–machine collaboration. The primary contribution of this research lies in proposing a Multi-Level Embedded Framework providing a complex view of AI-HRM in a manufacturing sector at micro, meso, and macro levels. The absence of sustainable HR transformation through AI integration was identified as the critical challenge at the macro level. This research provides theoretical and practical implications for designing the sustainable HRM system based on ESG principles and favors the United Nations Sustainable Development Goals 9 and 12.