The Delegation Tax: What AI Takes From the Mind That Uses It Too Much

A 2026 Harvard Business Review study found that 14% of workers using AI for professional tasks reported a phenomenon the researchers call 'AI brain fry' — a cluster of cognitive and motivational symptoms tied to over-delegation of mental work. Bedard, Kropp, Hsu, and Karaman found strong correlations between AI brain fry and information overload, reduced sense of competence, and motivational fatigue that sets in when the mind is perpetually catching up to outputs it did not generate.[^1] One in seven workers is not a fringe edge case.

The question this raises is not whether AI is useful. It plainly is. The question is what kind of use erodes the person doing the using.

What 'brain fry' actually describes

In Thomistic terms, the intellectual virtues — scientia, intellectus, prudentia — are formed through repeated acts of the intellect working under the will's direction. Aquinas argued in Summa Theologiae I-II that habits, including cognitive habits, are acquired and maintained by exercise.[^2] A person who consistently delegates the effortful portion of reasoning does not simply save time; they interrupt the loop through which competence compounds.

Benjamin Suazo's work on the cogitative sense is instructive here. The vis cogitativa — the highest of the internal senses, sometimes called the particular reason — connects universal principles to particular circumstances, enabling judgment in concrete situations. It is formed by attention, practice, and encounter with resistance. A tool that reliably smooths away the resistance is not neutral with respect to that faculty's development.

Jonathan Haidt's reading of the post-2012 cognitive data sits alongside the HBR findings usefully. Academic performance began declining well before COVID, coinciding with the period of heaviest smartphone and platform adoption.[^3] The HBR study extends that observation into adult professional life: the pattern is not confined to adolescent brains in formation; it appears in trained professionals who should, in theory, have stabilized intellectual habits.

The specific danger in academic settings

When AI use is encouraged in academic settings — and it increasingly is — the risks multiply because the setting is explicitly one of formation, not productivity. A student asked to write an essay is not being asked to produce text. They are being asked to think a problem through to a conclusion, notice where their argument fails, and revise. The essay is the trace of that cognitive labor. When AI produces the text and the student edits lightly, the labor is absent and so is the formation.

The Summa's treatment of prudence (II-II, qq. 47–56) identifies docilitas — teachability, the willingness to learn from experience — as one of prudence's integral parts. Docility requires that the learner encounter genuine difficulty and process it. An AI that converts difficulty into fluency before the student has engaged it bypasses the condition under which docility is exercised.[^2]

Grenny, Patterson, and McMillan describe a related mechanism in high-stakes conversation: scripted interaction reduces cognitive engagement to near zero, and the mind disengages.[^4] The student who submits an AI draft and 'edits' it is in something like this mode — following a script rather than authoring an argument.

Practical realities for people who cannot avoid AI

Abstention is not the answer. For many workers and students, AI tools are embedded in the systems they must use. The more useful question is how to use AI in ways that preserve rather than degrade the cognitive and moral capacities underlying good work.

Four distinctions matter in practice. First, AI amplifies when it handles operations the person already understands — formatting, consistency checks, summarizing documents one needs to contextualize but not master. It substitutes — and brain fry risk rises — when it performs the core reasoning operation the person is supposed to be developing. For a paralegal scanning case summaries, AI summarization amplifies. For a law student writing a brief, it substitutes, and the formation cost is real.

Second, intentional friction before prompting preserves the cognitive labor. Kevin Majeres, working within Catholic cognitive-behavioral frameworks, argues that attention is the ground of virtue formation. A practical counter to brain fry is drafting a rough answer in one's own words before prompting, then using AI to interrogate or extend that draft rather than originate it.

Third, Ignatius of Loyola's Rules for Discernment describe the importance of noticing interior movements as data about the orientation of one's actions. Workers who notice that AI use leaves them feeling hollowed out, passive, or oddly incompetent at tasks they once performed well are receiving a signal worth taking seriously. The HBR findings give empirical warrant to that interior data.

Fourth, prudence requires a fund of experiential knowledge built through repeated, unassisted engagement with difficult problems. For professionals and students who use AI heavily, scheduling protected time for unassisted work — writing, analysis, problem-solving — is maintenance of a cognitive infrastructure that AI use, at scale, is actively eroding.

The anthropological question underneath the empirical one

The CCMMP understands the person as a unity of body, soul, intellect, and will — created for rational and loving engagement with reality, fallen toward disorder in all those faculties, and redeemable through practices that re-form them toward their proper ends. AI brain fry is a specific instance of that fallen vulnerability: the path of least cognitive resistance is always available, the passions prefer fluency over effort, and the technology now offers something unprecedented — fluency without the work that makes fluency meaningful.

The redemptive path is not rejection of the tool but recovery of the agent. That means taking the HBR findings seriously not as a warning about a product but as a description of a person: someone whose cognitive habits are being shaped by a technology faster than they are shaping the technology's role in their life.

The 14% figure is a floor, not a ceiling. The researchers measured current, self-reported experience. The formation costs — the virtues not built, the prudential judgment not developed, the cogitative sense not exercised — will not show up in self-report surveys. They will show up, years from now, in the quality of judgment that experienced professionals and educated graduates bring to hard problems that no AI has seen before.

That is the argument for using these tools with intention and resistance, not with comfort and ease.

References

[^1]: Bedard, Kropp, Hsu, and Karaman, 'When Using AI Leads to Brain Fry,' Harvard Business Review (2026). [^2]: Thomas Aquinas, Summa Theologiae I-II (on habit and virtue) and II-II, qq. 47–56 (on prudence and its parts), trans. English Dominican Province (Westminster, MD: Christian Classics, 1981). [^3]: Jonathan Haidt, The Anxious Generation (2024). [^4]: Grenny, Patterson, McMillan, Crucial Conversations (2022).