AI AND ORGANIZATIONAL AGILITY: CAN AI OVERCOME LARMAN’S LAWS OF RESISTANCE?

Abstract

This study investigates whether the adoption of artificial intelligence (AI) in Agile organizations acts as a genuine catalyst of transformation or is co‑opted by existing structures, thereby reinforcing resistance patterns described by Larman’s Laws of Organizational Behavior. Building on a theory‑driven framework that integrates Agile transformation, organizational change, and human–AI collaboration, we analyze a survey of 97 practitioners across software, product and operations roles. The survey captures three constructs: level of AI usage, self‑reported satisfaction with AI tools, and confidence in the accuracy and reliability of AI‑generated outputs. Descriptive distributions indicate broad, but not yet deep, adoption: 79% use AI either for specific tasks or regularly with customization, whereas only 3% report deep, consistent integration. Satisfaction is high (~67% satisfied or very satisfied), while confidence is mostly moderate (54% moderately confident; 14% very confident). Exploratory associations suggest that higher AI usage and higher satisfaction are positively related to confidence in AI outputs. These patterns are consistent with an incremental adoption path in which AI is primarily applied to bounded, low‑risk tasks, avoiding disruption to decision rights and role boundaries—an empirical manifestation of Larman’s Laws. To explain these findings, we propose and visualize a conceptual model in which leadership and governance moderate the relationship between AI adoption and organizational agility: AI enables agility when accompanied by structural redesign and responsible governance, but risks becoming superficial when inserted into unchanged structures. The article contributes by (i) extending Larman’s Laws to the AI era with empirical evidence from agile settings; (ii) specifying measurable indicators for AI‑enabled agility; and (iii) outlining managerial implications that reconcile agile principles with responsible AI. We discuss limitations of the dataset and propose directions for longitudinal and multi‑method research on human–AI teaming, leadership, and structural change.