AI as Critic: Validating SAGE Pedagogy for Human Authority and Responsible GenAI Use in Systems Analysis and Design Education

Embedding Generative AI in higher education requires moving students beyond passive acceptance towards disciplined orchestration characterised by human authority over algorithmic suggestions. Prior SAGE framework implementations demonstrated reliable progression to balanced integration yet surfaced two constraints: a competency ceiling preventing theory-informed explanation, and an accessibility U-curve indicating non-functional awareness collapses when scaffolding is removed. This study completes longitudinal evaluation by following a multi-campus cohort (n=280) through testing-phase orchestration, operationalising a novel \emph{AI as Critic} model that inverts the power dynamic from students refining AI artefacts to defending human-created artefacts against algorithmic critique. Major Assessment-Project embedded adversarial review within an Apply-Defend-Synthesise sequence combining independent creation, principled acceptance or rejection of AI suggestions, and metacognitive synthesis. Qualitative analysis of six representative groups across five campuses confirms three advances. First, adversarial orchestration elicits contextual authority: students demonstrated selective adoption grounded in scope, platform constraints, risk prioritisation, regulatory compliance, and phasing strategy rather than passive deference. Second, Apply-Defend-Synthesise breaks the competency ceiling: students produced theory-informed critiques absent from AI summaries, evidencing Critical Synthesiser emergence. Third, the accessibility U-curve persists, confirming stimulus-bound competency requiring recurrent cueing rather than durable internalisation. Cross-disciplinary operationalisation demonstrates portability through instantiation in nursing, business, engineering, and education. Findings validate \emph{AI as Critic} as a reproducible mechanism for cultivating human authority and establish Apply-Defend-Synthesise as a transferable pattern for responsible AI orchestration with explicit guidance for embedding governance and non-functional qualities throughout curricula