Artificial intelligence (AI) supports manufacturing decisions involving revenue, suppliers, working capital, capacity and service commitments. Whether evidence from other sectors transfers to these decisions remains uncertain. This critical integrative review examines a reported corpus of 80 review-level publications dated 2015 to 31 August 2026, including a 59-review systematic or systematic-component subset. It distinguishes decision domain, AI decision role, evidence layer and manufacturing-primary scope, and integrates task-technology fit, organizational information processing, socio-technical systems and appropriate reliance. The reported classification identifies 11 manufacturing-primary, function-specific reviews: nine address production planning or scheduling, one demand forecasting and one inventory management. The corresponding restricted subset contains seven, one and one, respectively. These counts describe the selected corpus; they do not estimate field-wide prevalence or independent empirical support. The Manufacturing AI Decision-Fit (MAIDF) framework organizes evidence transfer around four regimes: data adequacy, feedback and actionability, operational coupling, and accountability. Five propositions and a non-compensatory decision workflow specify how these conditions may affect deployment value. The framework remains unvalidated, and the expanded corpus requires a complete record-level audit trail for independent reproduction. Its contribution is a structured account of transfer conditions across commercial decisions, with deployment assessed against a credible baseline and consequential business outcomes.