As far back as 1930, John Maynard Keynes imagined a future in which technological abundance would free humans from excessive labor, leaving space for leisure, art and human connection.
The Keynesian promise first emerged in factories. As machines automated physical labor, the question was whether workers would trade productivity for time or simply be asked to produce more. We know how that story unfolded.
But even as technology has accelerated, those productivity gains have rarely translated into time spent on better lives. It is a sobering reality: work has a way of expanding to fill whatever time remains.
Each new tool makes us faster, and each increase in speed quietly resets what counts as “enough.”
AI impact–more of the same
When it comes to productivity, artificial intelligence doesn’t introduce a new problem—it accelerates an old one.
What is striking about the AI moment, though, is that this same productivity dilemma has migrated upstream, appearing much earlier than many would have thought. The pressure no longer begins on the factory floor. The first place the productivity dilemma rears its head is not in industry, it is in the classroom.
Before students encounter AI-mediated workplaces, they encounter AI-mediated classrooms, manifested in syllabi, assignments, grading rubrics and informal cues about what “good work” looks like. When professors respond to AI by raising output expectations—more pages, more sources, more recommendations—we may believe we are preserving rigor.
In practice, however, we are actually training students to internalize a productivity reflex that never asks what the extra output is for. And that’s precisely the problem.
Reversing our instincts
Consider a familiar assignment: a short research-based essay or case analysis. Before AI, students struggled to synthesize sources, identify patterns and articulate recommendations.
With AI, much of that synthesis of sources happens faster. The knee-jerk reaction is to require more. Add sources. Expand the literature review. Require more appendices.
We make the assignment heavier to compensate for the tool rather than the other way around. That move may feel rigorous, but it is exactly how the productivity trap manifests and reinforces itself.
Across disciplines, the pattern is the same. We shrink the output and expand our judgment of work to give too much weight to sources and reference points.
The alternative is not to abandon standards, but to reverse the instinct. Instead of assigning more, we should under-assign, and shift what we assess and how. In a history course, this might mean fewer sources but greater responsibility for interpretation.
In a marketing class, it might mean asking for two strategic recommendations instead of five, and grading students on prioritization, trade-offs and live defense rather than volume. In an engineering school, it might mean focusing on why students made the assumptions they did rather than producing multiple models.
More optimized—and more exhausted
The history of productivity suggests that if we do nothing, we will graduate students who are faster, more optimized and more exhausted—without ever teaching them how to use reclaimed time for judgment, creativity or building meaningful relationships at work and in life.
This matters because the capacities we in higher education claim to develop—critical thinking, ethical reasoning, collaboration, leadership—do not scale with efficiency like coding or pumping out emails.
They require slack. They require presence. They require time to think. They require time that is not immediately converted into output.
We know the trap. The question now is whether educators are willing to resist it—deliberately and forcefully.
5 things can educators do
- Resist the temptation to ask for more output. If AI reduces the time required to complete an assignment, do not simply demand more volume. Same page count. Same scope. Higher expectations for thinking, not throughput.
- Design assignments that do not reward acceleration. Favor work that requires judgment, synthesis and reflection over sheer production. AI can generate drafts; it cannot replace discernment. Grade their decision-making processes, not just final artifacts.
- Deliberately under-assign. Leave space in the course by design. Name it openly. Tell students the absence of work is intentional, not an oversight. This is not free time. It is designed for conversation, integration and reflection.
- Shift rigor from quantity to maturity. Rigor does not have to mean more. It can mean slower feedback loops, deeper revision cycles or – better yet – sustained engagement over fewer ideas. Treat discernment as a desired skill that drives understanding.
- Use reclaimed time for relationship practice. If AI buys time anywhere, it should buy time for relationship development. Use class time for peer dialogue, disagreement navigation, collaborative sense-making and reflection. These are not inefficiencies. They are the work.
New relationship with productivity
Taken together, these moves generate something subtle but powerful: they model a different relationship to productivity. They teach students that efficiency can be good, but it is not a moral obligation; knowing when is enough is a critical component of intellectual adulthood.
AI will not deliver Keynes’s promised leisure on its own. The question is whether educators are willing to make a different choice at the point where norms are first learned.
If reclaimed time is always surrendered back to output, then the failure of the leisure society is not a mystery, it is a habit. And habits can be changed.
