<p>Artificial intelligence (AI) and machine learning are increasingly integrated into materials design and process optimization, yet their impact on additive manufacturing (AM) remains constrained by a fundamental and often overlooked bottleneck: tacit human expertise. Decisions that govern printability, defect mitigation, performance robustness, and qualification frequently rely on undocumented heuristics, experiential judgment, and contextual knowledge accumulated by practitioners. This Prospective defines AI-augmented materials design as the use of computational learning, mechanistic simulation, and human-guided optimization to support materials decisions across the AM workflow. It further defines AI-augmented materials knowledge systems as socio-technical infrastructures that organize multi-modal data, thermomechanical simulations, and explicit representations of expert intent into reusable decision-support resources. We examine where tacit knowledge dominates AM practice, identify limitations of data-centric AI approaches, and propose a layered framework that connects sensor data, CFD/FEA-informed simulation outputs, expert annotations, and human-in-the-loop learning. Through vignettes in metal AM, polymer/composite AM, and printed functional materials, we show how such systems could improve transferability, reduce trial-and-error, and support scalable qualification while preserving expert judgment.</p> Graphical abstract <p></p>