Every HR function I talk to is trying to answer the same question this year: how fast can we get managers using generative AI. A new University of Bath-led study says that is the wrong question, and answering it well could make managers worse at their jobs, not better. If HR’s AI rollout plans do not account for that risk, they are optimizing for adoption speed while quietly eroding the judgment the organization actually depends on.
The case for going faster anyway
The obvious counter-argument deserves stating plainly, because it is not wrong on its own terms: generative AI genuinely makes managers faster. It drafts the performance review, summarizes the market research, models the headcount scenario, and frees up hours that a time-pressed manager can spend on the parts of the job that supposedly matter more, coaching people, resolving conflict, thinking strategically. On that logic, the manager who resists AI tools is the one falling behind, and HR’s job is to remove friction from adoption as quickly as possible.
What the research actually found
The counter-argument assumes the hours AI frees up get reinvested in judgment. The research, published in the Academy of Management Review by a team from the Universities of Bath, Ohio, Lausanne and Cardiff, says that assumption is exactly backward for managers under time pressure, which is to say most managers. The team’s concept is “epistemic de-skilling”: the gradual loss of knowledge-related capability that happens when people outsource too much thinking to generative AI rather than working through it themselves.
“Gen-AI appeals because it can help people complete tasks more quickly. However, Gen-AI cannot replace the lessons learned through first-hand experience. Unlike humans, AI does not experience the world, understand the consequences of decisions, or grasp the social and emotional complexities in our workplaces. Instead, it produces responses based on patterns found in existing data,” said Professor Dirk Lindebaum of the University of Bath’s School of Management, who led the research.
The mechanism the researchers describe is specific: managers stop asking hard questions, stop seeking different perspectives, stop learning from the messy real-world interaction that used to force them to build judgment in the first place. “In these situations, managers may stop asking important questions, seeking different perspectives or learning from real-world interactions. Instead of developing a nuanced understanding of employees, customers or organisational challenges, they may come to depend on AI-generated answers that lack the context and moral judgement needed for complex decisions,” Lindebaum said. The risk is highest, the study found, exactly when time pressure is highest, which describes most of the managers HR is currently pushing to adopt AI fastest.
The fix is not less AI, it is a different design
None of this is an argument for keeping generative AI away from managers. The researchers found the same tool can build judgment instead of eroding it, through what they call “epistemic up-skilling”: using AI outputs as a starting point to interrogate, not a finished answer to accept. “Rather than accepting AI outputs at face value, managers can use them to challenge assumptions, explore alternative scenarios and test the reasoning behind their own decisions,” Lindebaum said, adding that the approach works best when managers know they will be held accountable for explaining their reasoning to someone else. “It is that which Gen-AI cannot satisfactorily explain that managers must explain to themselves and others.”
That is a workflow design problem, and workflow design is HR’s job, not the AI vendor’s. A rollout that hands managers a chatbot and a productivity target will get the de-skilling outcome, because nothing in that design asks the manager to do the harder work of scrutinizing the output. A rollout that pairs the same tool with real accountability, requiring managers to explain and defend the reasoning behind an AI-assisted decision, gets a different outcome from the same technology. This is the same distinction already playing out in how AI is displacing the informal mentoring relationships that used to build this judgment implicitly, and it echoes the broader caution that vendor productivity claims deserve more scrutiny than they usually get before they reshape how a function actually works.
What this means for the HR leader
The practical takeaway is not to slow down AI adoption among managers. It is to stop measuring the rollout by adoption rate alone. Lindebaum’s team is blunt on this point: “It is becoming increasingly clear that simply introducing AI tools will not automatically improve decision-making or organisational performance. Instead, organisations need to carefully design roles, responsibilities and workflows to ensure employees continue developing the human skills that AI cannot replicate.” HR functions that treat AI training as a one-time enablement session, rather than an ongoing redesign of how managers are held accountable for their reasoning, are building the de-skilling risk into the program by default.
Concretely, that means building review and calibration processes that ask managers to articulate why an AI-assisted recommendation is right, not just whether they used the tool. It means protecting the accountability structures, real performance conversations, real post-mortems, real challenge from peers, that force a manager to reason rather than defer. And it means treating “our managers use AI heavily” as a metric that says nothing on its own about whether the organization’s judgment is getting better or quietly hollowing out.
Source: University of Bath
