The rapid integration of artificial intelligence into digital learning environments has surfaced a persistent and consequential challenge: quality assurance frameworks are evolving far more slowly than the technologies they are meant to govern. This experience paper reports on two complementary inquiries illuminating quality assurance in AI-enabled digital learning at two distinct but interdependent levels. The first, conducted through a multimodal engagement process with the Quality Matters AI Working Group culminating in a design sprint at the QM Board Retreat in November 2025, examined quality assurance at the system level, exploring how organizations and institutions might govern, guide, and continuously improve AI adoption in ways that preserve educational integrity. The second, conducted in partnership with D2L between November and December 2025, examined quality assurance at the artifact level, investigating the practical challenges of evaluating AI-generated educational content and where current evaluation frameworks fall short. Together, these bodies of evidence converge on a shared insight: quality assurance in AI-enabled environments cannot be reduced to checklists, automated scores, or static rubrics. These two levels are structurally interdependent. Artifact-level evaluation without system-level governance produces inconsistent and acontextual judgments, while system-level governance without artifact-level evidence produces policy that is principled but untethered from practice. Quality assurance, understood this way, is not only a curriculum problem. It is an ecosystem problem. Drawing on human-centered AI frameworks (Shneiderman, 2022), TPACK-informed perspectives on educator knowledge (Koehler & Mishra, 2009), established peer-review approaches to online course quality (Legon & Garrett, 2018), and Hau's (2025) concept of Relational Quotient, the paper argues that the relational dimensions of quality assurance are not the soft periphery of QA practice but its load-bearing structure. Hau proposes that the distinctive human capacity to connect, collaborate, and build trust becomes more essential, not less, as AI manages more of the content layer of learning.