Loading today’s stories
Science1 distinct publisher3 min readPublished
A new Academy of Management Review paper argues that routing thinking through Gen-AI under time pressure costs managers the practical wisdom only firsthand consequences build. The argument is conceptual, and its two conditions are testable.
The Scientist · Science desk
Compiled by The ScientistSomething wrong?How this is made
Practical wisdom, in this paper's sense, is an accumulation rather than a trait: the moral reading, the contextual grasp and the know-how a manager picks up from real situations, from reflecting on them and from dealing with other people [1]. Dirk Lindebaum's argument is not that the output of a chatbot is poor. It is that the system has no stake in what follows from its answer, because it does not experience the world or bear the consequences of a decision, and instead assembles responses from patterns in existing data [6]. Route idea generation or a live problem through it and the memo may be perfectly serviceable while the manager's own learning loop goes quiet. Over time, the paper argues, the capacity to learn from experience and to anticipate what future goals will require thins out [7].
The article's title is the honest guide to its evidentiary weight. It is a process model [2], written by researchers at Bath, Ohio State, Cardiff and Lausanne [3], and the reported work carries no cohort of managers and no comparison group, so there is no effect size in it [12]. For a theory paper that is not a defect. It does bound the use: what you get is a mechanism and two switches, not a rate of decay.
Both switches sit in the workplace rather than in the tool [13]. Intense time pressure, with Gen-AI used as a shortcut instead of engagement with the problem, is the condition under which the authors expect epistemic deskilling [4][5]. Accountability is the condition they expect to run the other way, where managers must justify their actions and reasoning and the tool becomes a prompt for reflection rather than a replacement for it [10]. The mechanism there is oddly specific and worth keeping: because these systems often cannot say why they produced a particular answer, the unexplained part is left for the manager to explain, which takes persistent effort [9]. The authors say this upskilling path is the harder one to obtain [8].
Notice what that remedy costs. Filling explanatory gaps takes time, and the trigger the paper flags for deskilling is the shortage of time [14]. An organisation that drops Gen-AI into a workflow and leaves the deadlines exactly where they were has, on this model, installed the risk condition and skipped the countermeasure. Which is why Lindebaum's own prescription is about roles, responsibilities and workflows rather than about which tool to buy [11].
The thing this does not tell you is the dose, whether the loss reverses, or whether managers who have thinned out in this way make measurably worse decisions. A process model cannot answer those [12]. The narrow claim I would still put weight on is the asymmetry between the two conditions: time pressure is the default state of most managerial work, and a requirement to explain your reasoning is cheap to impose, so the deployment this model predicts worst is the one that adds the tool and changes no review obligation at all.
Ranked by verification strength, evidence, and original report placement.
A study from the University of Bath, published in the Academy of Management Review, explores how tools such as ChatGPT may affect "managerial phronesis", the practical wisdom managers develop through real-world experience, reflection and human interaction; overreliance may erode managers' ability to build moral insights, contextual understanding and know-how to get a job done.
The paper is Dirk Lindebaum et al, "A Process Model of Managerial Phronesis in the Age of Generative AI", Academy of Management Review (2026), DOI 10.5465/amr.2024.0582.
The research team comprised Professor Dirk Lindebaum of the University of Bath's School of Management, Professor Natarajan Balasubramanian of Ohio State University, Dr Mehreen Ashraf of Cardiff University, and Dr Patrick Haack of the University of Lausanne.
The team identified the concept of "epistemic deskilling", a process in which people gradually lose knowledge-related capabilities because they outsource too much thinking to Gen-AI.
The researchers suggested epistemic deskilling was most likely to happen when managers are under intense time pressure and use Gen-AI as a shortcut rather than engaging deeply with a problem themselves.
Lindebaum said that unlike humans, AI does not experience the world, understand the consequences of decisions or grasp the social and emotional complexities in workplaces, and instead produces responses based on patterns found in existing data.
Distinct publishers with included, body-backed reporting in this cluster.
1 article · September 5, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
build
Thirty runs of one prompt put a 34-point error bar around a brand's mention rate1 distinct publisher
science
Text watermarks land on 2 December. The detection they imply does not.1 distinct publisher
invest
A Connecticut judge just priced prompt injection: no fine, no e-filing2 distinct publishers
product
OpenAI's plan to hand everyone a coding agent leaves the hard part to the model1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One press write-up of a theory paper
The provenance is easy to verify: a named journal, a DOI, four identified authors at four universities. The claim itself is not verifiable from this reporting, because a process model is an argument about mechanism and Bath's account carries no measurement of any manager's judgment before or after Gen-AI use. Academy of Management Review is a serious venue for theory-building work, which is what holds this in the middle of the range rather than at the bottom.
Nothing to count yet
A publication date is not uptake. No organisation in this reporting has redesigned a role, changed a deadline policy or measured manager behaviour in response, and there is no usage figure for the kind of reflective prompting the authors describe, so there is nothing here to score.
A model reported as a warning
The hedges are in the right places: "could undermine", "may erode", and Lindebaum's own closing line concedes that tools alone change nothing. What a reader is never told is that the warning rests on a model. The word "study" does the work of implying evidence, and only the citation at the bottom reveals that the output is a process model. The distance between the register of the write-up and the standing of the underlying claim is the whole of the overstatement.
University communications, carried onward
Bath's press operation had one purpose here and Phys.org's republishing model fits it neatly: quotes preserved, mechanism explained, no outside scholar asked whether the model survives contact with a working management team. Lindebaum is the only voice in the piece. That does not make the argument weak, but nobody in the chain was positioned to ask for the data the argument lacks.
Solid on the citation, thin on the effect
Who wrote what, where it appeared and what it asserts are all firm. Frequency and magnitude are not: how often deadline pressure actually produces the deskilling described here is untouched by this reporting, and with one account of one paper there is no second reading to triangulate against.