AI Overuse Quietly Erodes Managerial Judgment Skills
A quiet shift has taken place in offices worldwide. Generative AI now drafts emails, summarizes meetings, and proposes strategies before a manager has fully formed a thought. The convenience is real. So is the cost, according to new research from the University of Bath. The study, led by researchers at the University of Bath’s School of Management, warns that routine reliance on generative AI slowly strips away three core managerial abilities: moral insight, contextual understanding, and practical know-how The tool that was meant to amplify judgment may instead be quietly replacing it.
The finding lands at a peculiar moment in workplace history. For decades, management theory held that experience builds judgment through repeated exposure to messy, real-world problems. Each difficult decision added a layer of nuance to a manager’s mental model. Each mistake taught something no textbook could. The Bath study suggests generative AI interrupts this process at its root. When the machine supplies the answer, the manager never exercises the muscle that builds the insight. The skill does not disappear overnight. It atrophies, like an unused limb.
The research draws a stark comparison with earlier technological shifts in the workplace. Spreadsheets automated arithmetic, yet managers still needed to understand the numbers behind them. Search engines made facts instant, yet analytical thinking remained the bottleneck. Gen-AI is different in kind, not just in degree. It does not automate a discrete task. It automates the reasoning itself. Earlier tools left the thinking to the human and sped up the execution. This one offers to do both, and the offer is difficult to refuse when deadlines press and inboxes overflow.
The core mechanism identified by the study is overreliance. A manager who checks an AI-generated analysis as a starting point behaves differently from one who accepts it as the final word. The former retains a critical stance. The latter enters a state the researchers describe as passive acceptance. Over time, this passivity becomes a habit. The manager stops asking why a recommendation makes sense because the machine always provides a coherent answer. Coherence, however, is not the same as correctness, and fluency is not the same as understanding.
The study names what is at stake. Moral insight, the researchers argue, develops through grappling with ethical ambiguity. A layoff decision, a resource allocation conflict, a performance review with contradictory evidence — these situations resist algorithmic resolution because they require weighing values against each other. Contextual understanding follows a similar path. Every organization has unwritten rules, historical tensions, and personal dynamics that shape what a good decision looks like. No training dataset captures those specifics. Know-how, the third eroded capacity, is the most practical: the tacit ability to get things done through informal channels, timing, and relationship awareness.
These three capacities are not taught in isolation. They emerge from the practice of making decisions and living with the consequences. The study implies a troubling trade-off. Every hour saved by delegating a judgment call to Gen-AI is an hour not spent developing the judgment needed for the next, harder call. The efficiency gain is immediate and measurable. The loss is slow and invisible, accumulating until a manager faces a situation where the AI has no answer and discovers the internal resources have dwindled.
The researchers do not argue for abandoning the technology. Their position is more nuanced: the problem lies in how the tool is used, not in its existence. A manager who treats Gen-AI as a thinking partner rather than an oracle preserves the cognitive work that builds expertise. The distinction sounds simple in theory. In practice, it demands constant vigilance against the seduction of convenience. Every prompt is a choice about whether to delegate judgment or exercise it.
The study opens a question that demands further measurement. If overreliance erodes judgment, at what pace does the erosion occur? Is it a matter of months or years? Does the damage reverse when usage stops, or does it leave a permanent mark on cognitive habits? The University of Bath research establishes the phenomenon. The next step, the researchers suggest, is quantifying its trajectory so organizations can design safeguards before the cost becomes visible in their leadership ranks.