RungBuilder™: From AI-Assisted Performance to Accountable Human Capability

RungBuilder™ is an open developmental and entrustment architecture for building, demonstrating, and responsibly relying on human capability in AI-assisted work. Generative AI can improve the quality and speed of work while weakening the historical relationship between a polished work product and the human capability it appears to represent. In institutions, it can also compress or remove the formative work through which novices develop expertise, judgment, and the ability to supervise increasingly capable systems. RungBuilder addresses these related problems by helping institutions reconstruct developmental pathways from assisted participation to evidence-backed, bounded responsibility. The architecture integrates established foundations in cognitive apprenticeship, scaffolding, deliberate practice, transfer, competency-based education, Entrustable Professional Activities, and programmatic assessment with emerging approaches to AI fluency and AI-era assessment reform. Its core elements include Shadow → Steer → Own, the eight-step Accountable Learning Episode, Learning Integrity, selective evidence of consequential human judgment, transfer testing, evidence-credibility checks, and bounded entrustment. RungBuilder is not an LMS, AI product, credentialing authority, or claim to have originated the problem of automation-induced skill loss. It is a practical reference architecture that universities, employers, professional firms, healthcare educators, workforce programs, and public institutions can adapt and test using systems they already possess. This concept paper presents the architecture, its theoretical foundations, implementation patterns, governance considerations, pilot methodology, and research agenda. Its effectiveness remains an empirical question to be tested through practical implementation.