Work vs. gym: when to use AI, when to skip it

Bruce Schneier, a security technologist teaching at Harvard Kennedy School and the Munk School at the University of Toronto, has spent years observing his students use AI on writing assignments—and has developed a framework for thinking about when AI assistance makes sense and when it doesn't.

His key insight draws on a distinction made by AI researcher Daniel Meissler: the difference between work and the gym. If your job is to move heavy things from one side of a room to the other, use whatever assistive technology you have—a wagon, a forklift, an AI robot. But at the gym, it makes no sense for a robot to lift weights for you, because the point of weightlifting is not to move heavy things; it's to actually lift them yourself, to build strength. The same logic applies to any task AI can do. If the task is work—if it has to be done and no one cares how—then AI assistance is appropriate. But if the task is more like the gym, where how you do it is as important as the outcome, then using AI probably does not make sense.

Schneier emphasizes that this framework assumes the AI is actually trustworthy: that it can perform the task well, that its mistakes are minimal and correctable, and that it is secure from cyber-attacks. Once you are confident the AI can perform reliably, the work-versus-gym distinction guides the decision to use it. His student writing assignments are gym tasks, not work tasks. He assigns them because the act of writing—thinking, outlining, drafting, editing, making and criticizing and revising arguments—develops critical thinking skills students will need in their careers. Without this constant mental exercise, those skills atrophy. Reading the assignments students turn in, Schneier can see whether their skills are flourishing or atrophying; he can usually tell the difference between an AI-written memo and a student-written one, especially when the student simply turns in what the chatbot produces. AI-generated writing in mid-2026 tends to be catchy, plausible, and grammatically perfect, yet not particularly well-crafted or logically coherent.

What makes Schneier able to spot this is his own years of developing writing skills. Students, by contrast, do not yet have that skill; they mistakenly view a confident, well-written essay as evidence of the quality of their ideas. They see AI as cleaning up their ideas and helping them get past the uncomfortable stretch of turning those ideas into prose. What they miss is that their initial discomfort is a normal and healthy stage of writing, not something to quickly skip past. The struggle to express what they think is an important part of the process—it is how they test their ideas, examine their hypotheses, and actually figure out what they think.

The work-versus-gym distinction also illuminates the problem facing creatives. Most of the time someone hires a writer, they just need the words: an instruction manual, a detailed sales presentation, a government-mandated disclosure document, a legal brief—dry, predictable, accurate writing. That is work, exactly what AIs are good at today and what Schneier does not want in student assignments. Only sometimes is writing an art form—a book, a poem, an uplifting political speech—where process matters as much as product. For most of human history, the only option for all these tasks was human writers, regardless of whether the task needed work writing or gym writing. That employment paid many writers' salaries. Now, for the first time in human history, society can separate work-type writing from gym-type writing. If AI can do most work-type writing, society does not need as many human writers.

The same holds for visual artists. Sometimes an actual artist is needed, but most of the time a simple image suffices: a corporate mascot, a warning sign, a packaging label. Historically those jobs went to artists, and sometimes beautiful art resulted. But most of the time it was just work. The world needs less pure art than simple images.

Yet explaining the problem is not the same as solving it. Schneier gives his students the work-versus-gym speech every class, but they still use AI. He has sympathy: assignments are hard, everyone is overworked and overstressed, and—most importantly—students feel they will look bad if their peers are all using AI. Even if they do not want to use the technology, they feel they have no choice. There is also an incentive problem. No one pays people to go to the gym; maintaining healthy habits requires discipline. The payoffs to exercise—fewer aches and pains, less fatigue, better mood and stress management—might make Schneier a better writer and teacher, but they are subtle and easy to miss. For students, incremental improvements in reasoning and writing are equally subtle.

Schneier argues that we do have a choice. We can look at the tasks of our lives and separate them into work or gym. Just as we might choose stairs instead of an elevator or walking instead of calling an Uber, we can wall off cognitive gym tasks from AI and ensure we do not lose our skills. The same applies when assigning a job to someone else: if it is a work task, AI can do it; if it is a gym task, it is a waste of everyone's time to give it to AI because no one learns or gets stronger.

A future where AI generates words and images is one where society must make deliberate choices about how it treats its creatives. This will not be the first time—today there is minimal demand for portrait painters—but perhaps this time society can make different, more deliberate choices about the value of art. AI will fundamentally change the nature of work, though not nearly as fast as AI companies want people to believe. Policy analysis will definitely involve AI from now on, and students need to reimagine what it means to learn and practice that skill. The line between work and gym will change in the future as humans adapt to a world with these new intelligences. But for now, Schneier argues, the distinction is pretty clear: use it on yourself to decide.

Schneier's framework addresses a genuine tension in the age of capable AI: the temptation to optimize every task by outsourcing it to a tool. His work-versus-gym distinction is grounded in a practical observation from his classroom—students turn in grammatically perfect, confident essays that lack logical coherence and intellectual rigor. Schneier attributes this to the fact that students, lacking years of writing practice, cannot yet tell the difference between polished prose and sound reasoning. They mistake AI output for a sign that their ideas are solid, when in fact they are skipping the messy, essential step where raw thinking becomes testable argument.

The framework extends naturally to creative professions, where he notes that most hired writing and art has historically been functional rather than artistic. Technical writing, legal briefs, packaging design, and portraiture were never primarily valued for artistic merit; they were work, and AI is simply better equipped to do work-type tasks cheaply and reliably. What changes is that society can now, for the first time, separate those two demands—paying for art only when art is actually needed, rather than hiring artists for every task. Schneier acknowledges this will disrupt creative careers, but frames it as an opportunity for deliberate policy choices rather than inevitable decline.

He also surfaces an incentive problem: maintaining cognitive skills through practice (the gym) offers subtle, delayed payoffs—better reasoning, fewer mental aches—while outsourcing offers immediate relief. Students feel pressure to use AI if peers are using it, and the long-term cost of skill atrophy is easy to overlook. Schneier's solution is not to ban AI, but to consciously choose where to use it and where to abstain, just as someone might choose stairs over an elevator. He notes that this line between work and gym will shift as humans and AI co-adapt, and that professionals like policy analysts will need to reimagine what learning their skill means in an AI-assisted world.