Who stands behind this?

Different articles. Two different fields. Two different decades.¹

AI can perform tasks, make recommendations, and even appear to exercise judgment. But it cannot take responsibility for the consequences of what it does. That responsibility remains with the people and organizations deploying it.

Philosophy of technology identified this as the "responsibility gap" two decades ago. Management research is now arriving at much the same conclusion through concerns about phronesis and deskilling.

Delegating a decision to AI does not delegate accountability. If anything, the more capable the system becomes, the easier it is to mistake successful performance for responsible judgment.

Two different fields. Two different decades. Neither was talking to the other.

AI might take work off your desk. It cannot take responsibility off your hands.

—

Someone once said that the factory took the hand from the craftsman and gave it a lever; the worker learned one motion, and once learned, it asked nothing more of him than consistency. Muscle memory mattered. Intelligence didn't. That someone is Karl Marx and he called it alienation.

Something similar is occurring with AI, but we shouldn't rush to a conclusion.

Fast version: AI is only the next machine doing the same thing to the next kind of hand, except this time the hand is judgment and the factory floor is the office.

Harder version: AI is a system trained on an ocean of prior human output, producing something that looks, from a reasonable distance, like judgment.

There is a however. Actually, there are two.

First, what looks like judgment is recombination. Every output is built from sedimented human judgments already sitting in the training data. Recombined, not synthesized the way a person synthesizes, where someone who has lived somewhere, decided something, and stood behind a conclusion puts it forward as theirs. The machine has no somewhere. It has a corpus.

Second, accountability doesn't travel with the recombination. When a manager makes a judgment call, someone makes it (a "who" rather than a "what"): a person who can be asked why and who bears what follows. That persisting, answerable subject isn't a nicety attached to judgment afterward. It's what makes judgment judgment rather than output. Take the person away, and what's left might be accurate, even useful. It isn't judgment. It's a very good guess in judgment costume.

So, two papers, two fields, twenty years apart, neither reading the other, landing in the same spot anyway: a system can perform the function without holding the position. Nobody coordinated that convergence. But somebody stood behind each paper. Matthias put his name to an argument, alone. Lindebaum and his co-authors did the same, two decades later, in a different discipline. The through-line isn't citation. It's accountability. And that's the very thing both conclude AI doesn't have.

So the inherited story (man loses his hand to the machine) turns out to be only half of it. The other half is that the tasks judgment used to require were defined, quietly, to exclude judgment in the first place. They were reduced to something automatable because somewhere upstream, someone had already stopped asking who stands behind this.

AI didn't do that. It just made the absence visible.

Marx would be proud.

¹ Dirk Lindebaum, Natarajan Balasubramanian, Mehreen Ashraf & Patrick Haack, "A Process Model of Managerial Phronesis in the Age of Generative AI," Academy of Management Review (2026). Andreas Matthias, "The Responsibility Gap: Ascribing Responsibility for the Actions of Learning Automata," Ethics and Information Technology (2004).

==

Housekeeping: Management as Craft now lives on my website, not Substack. You will be receiving emails from rich-bri@ghost.io. Add it to your contacts if you want to be sure it keeps landing.