Organisational instability and technology inconsistency make professional learning a significant yet often neglected challenge. Studies on digital transformation have uncovered a consistent decoupling between technologies and learning capacity (Vial, 2019).

Objective. Research aims to develop, validate, and psychometrically operationalise an original conceptual framework, the RAPS model (RAPS) – Resilient Agency through Psychological Safety. The model demonstrates how adult professionals maintain ongoing learning in an AI-saturated organisational space with structural ambiguity.

The RAPS framework integrates six constructs that are derived from theories, namely, organisational learning culture, psychological safety, Stoic resilience, learning dispositions, adaptive performance, and human-AI trust. According to Edmondson and Lei’s (2014) construct history and meta-analytic evidence pertaining to contextual variance (Frazier et al., 2017; Newman et al., 2017), RAPS views resilient agency as a digitally mediated and socially enacted capacity rather than a stable trait.

Methodology. The research applies instrument development mixed-methods procedures (Creswell and Plano Clark, 2018).

Findings. Phase 1 findings show that expert panels from two cultural settings independently challenged two prevailing design assumptions. First, psychological safety should not be presumed stable across organisational hierarchies in AI-saturated environments. Power proximity and algorithmic surveillance function as structural antecedents not addressed by team-level models, consequently extending the research agenda outlined by Edmondson and Bransby (2023). Second, panelists consistently identified human-AI trust as a probable predictor of psychological safety rather than its downstream consequence. They contended that distrust in organisational AI governance inhibits learning disclosure before interpersonal climate variables become relevant. Although the WEF (2023) classifies AI transparency as a governance variable, RAPS expert data repositions it as a design variable for learning environments.

These propositions challenge two enduring assumptions in digital learning assessment and invite critical dialogue regarding construct validity, cross-cultural measurement, and the governance conditions that support professional learning under structural uncertainty (Kegan & Lahey, 2016).