GenAI and the deskilling effect
Academic research into how Generative AI transforms workplace structures has uncovered counterintuitive effects on management hierarchies, challenging assumptions about AI-driven organisational flattening.
The study by University of Connecticut and Fudan University researchers examined how GenAI's dual nature –enhancing worker capabilities whilst introducing hallucination risks– affects organisational design. They found that firms should only adopt GenAI when error rates fall below specific thresholds, beyond which potential losses outweigh managerial cost savings.
A key finding contradicts conventional wisdom: as GenAI becomes more capable or reliable, organisations don't necessarily need fewer managers. Instead, companies strategically reduce worker skill requirements to cut wage costs, but these less experienced employees then require more oversight. This creates what researchers term an organisation “deskilling effect" that can increase management needs.
The validation dilemma presents another paradox. When firms implement manager review processes to catch AI errors, expensive validation procedures actually force companies to hire more skilled workers who need less supervision—ultimately reducing manager requirements despite adding oversight tasks.
The study develops a theoretical model to examine GenAI’s impact on organisational structure and the role of human-in-the-loop oversight
The research identified three distinct productivity phases: initial workforce reduction with declining worker income, followed by a "productivity bandwagon" phase where individual wage increases offset job losses, and finally a transition to flattened organisations of highly skilled autonomous workers.
These findings suggest GenAI's organisational impact depends heavily on implementation choices around worker skill levels, validation processes and productivity targets rather than the technology's capabilities alone.