Should You Use AI for a Task? Here's a Simple Way to Decide

Bruce Schneier, who teaches public policy at the Harvard Kennedy School and the Munk School at the University of Toronto, admits his students lean on AI to knock out writing assignments constantly. He also thinks it's a waste of their tuition. Those two facts sit uneasily together, and his essay tries to resolve the tension with a borrowed idea from AI researcher Daniel Meissler: think of tasks as either work or gym.

The distinction is blunt but useful. If you're hauling heavy boxes across a room for your job, grab a forklift, a wagon, an AI-powered robot, whatever gets it done — nobody cares how the boxes moved, only that they did. But if you're at the gym lifting weights, having a robot do the lifting defeats the entire point. The exertion is the product. Schneier argues writing assignments, especially policy memos, are gym tasks: he doesn't need more memos in the world, he needs students who've wrestled with outlining, drafting, arguing, and revising until their reasoning sharpens.

He's candid that he can usually spot the difference between an AI-written memo and a student-written one — the AI version reads confident, grammatically clean, and hollow underneath, missing coherent logic. His students, though, often can't see that gap. They mistake polished prose for good thinking and treat the AI as a tool for smoothing over the uncomfortable stage of turning half-formed ideas into sentences. Schneier says that discomfort is exactly the part they need to sit with, because struggling to articulate a thought is how you figure out whether you actually believe it.

The framework scales beyond the classroom. Most professional writing — manuals, disclosures, legal briefs — is work-type writing where AI assistance makes obvious sense. Only some writing, like books, poems, or a speech meant to move people, is gym-type, where process and product carry equal weight. Historically, human writers did both kinds because that's all there was, and Schneier notes some fiction writers financed their art with dull technical writing gigs. Now that AI can split the two apart, he argues society needs fewer human writers for work-type output — and the same logic applies to visual artists producing mascots, signage, or packaging rather than actual art.

Schneier doesn't pretend the fix is simple. He gives the work-versus-gym pitch every class, and students keep using AI anyway, partly from stress and partly from fear of falling behind peers who are also cutting corners. There's also no built-in incentive to stay disciplined, the same reason people skip actual gyms — the benefits are real but slow and easy to dismiss. Still, he insists individuals and institutions retain the choice to wall off gym tasks from AI, the way someone might take the stairs instead of the elevator, so the underlying skills don't quietly erode.

He closes by conceding that AI will reshape work eventually, just slower than AI companies claim, and that the boundary between work and gym tasks will keep shifting as people adapt. But for right now, he says the line is clear enough to use on yourself.