What Do We Do With Our Time When AI Does All the Work?

“The unexamined life is not worth living.” — Socrates, 399 BC

There is a question that I find the artificial intelligence industry reluctant to sit with, at least in my experience building within it. Not because the people building these systems are incurious or careless. Many of them are thoughtful, and some are genuinely brilliant. The discomfort runs deeper than that. The question touches something the industry’s operating logic tends to leave unexamined.

The question is this: if AI eventually does most of the work, what do we do with our time?

It sounds simple. It is not. Buried inside that question is a tangle of assumptions about what work actually is, what it does to us, and whether the thing that replaces it, whatever that turns out to be, can do the same job. The economic question (how we distribute resources, restructure labor markets, and compensate people whose skills are rendered obsolete) is genuinely difficult. Entire industries will be displaced. Entire systems of incentive and value will need to be rebuilt. And underneath the economic question is a human one that cannot be separated from it: what are people for, once the work is done? Lose the job and you lose the income, but you also lose the identity, the structure, the sense that your contribution matters. The economic and the human are not two separate problems. They are two faces of the same crisis. I want to examine both, and the place where they meet.

I build AI systems for a living. I have spent years inside the machinery of this transition, helping companies integrate intelligent automation into products their customers depend on. I am an operator, a builder, someone with skin in the game of what AI becomes. And from that position, I have grown increasingly unsettled by how few people inside this industry are asking the question seriously.

So I am asking it here. Not as a thought experiment. The displacement is already underway, and so is the resistance to it. Without false resolution. Without the optimist’s lazy comfort or the pessimist’s lazy dread. Just the question, held carefully, examined honestly.

I. The Original Affluent Society

To understand where we might be going, it helps to understand where we have already been. And the historical record, if you look carefully, contains a startling fact: for most of human prehistory, we did not work the way we think of work today.

We assume the forty-hour week, the career arc, the lifelong obligation to produce. We assume all of this is natural, baked into what it means to be human. But before agriculture, before cities, before any of the structures we now take for granted, our species lived a very different way. Hunter-gatherer societies, the mode of life that defined our species for roughly 95 percent of our existence, required approximately fifteen to twenty hours per week devoted to food acquisition.¹ The anthropologist Marshall Sahlins, studying these societies in the 1960s and 1970s, coined a provocative phrase for what he found: the original affluent society. Not affluent in the sense of material abundance. Affluent in the sense that their wants were modest and their means were sufficient. The rest of the time, the majority of waking life, was spent on story, ritual, play, relationship, and rest. Total work hours including food preparation, tool-making, and domestic tasks were higher, but the broader point holds: foragers appear to have worked considerably fewer hours than agriculturalists and industrialists.²

Then agriculture happened. And agriculture, despite everything civilization owes to it, made us work more, not less. We traded leisure for surplus, security, and the explosive population growth that eventually produced everything from written language to the smartphone in your pocket.

But this was not a trade freely chosen by most of humanity. Agricultural societies did not replace foraging societies because farming was a better way to live. They replaced them because farming produced larger populations, greater surpluses, and more military power. The societies that planted grain could field armies that the ones still hunting could not match. Industrialization repeated the pattern: industrial nations did not merely outperform agrarian ones; they dominated them. The narrative of progress, from foraging to farming to factories, is the narrative written by the civilizations that won. The losers of these transitions rarely got to tell their version.³

This matters for the question at hand. The AI transition may follow the same logic. The companies, industries, and nations that adopt AI most aggressively may not be building better lives for the people inside them, but they will outcompete the ones that do not. That competitive pressure, not any reasoned assessment of human flourishing, may be what determines the pace and depth of the transition. It was not a reasoned assessment that ended the foraging way of life. It was ten thousand years of being outbred and outfought by people who worked harder and had more grain.

From the moment the first grain was planted, “idle hands” became a moral problem to be solved. Religion, culture, and later the full apparatus of industrial capitalism built elaborate systems over thousands of years to ensure that people stayed productive.

The point is not that hunter-gatherer life was idyllic. It was dangerous, short, and often brutal. The point is that the assumption built into every conversation about AI and employment, that humans have always worked this hard and always will, is historically false. We once had the free time. We filled it with meaning-making. What we do not know is whether modern humans, stripped of that orientation over a hundred generations of agricultural and industrial conditioning, can find their way back to it.

II. What Rome Already Taught Us

We do not have to speculate entirely. History has already run a version of this experiment, and the results are instructive, and sobering.

At the height of the Roman Empire, a significant portion of the Italian peninsula’s population were enslaved. Scholarly estimates range widely, from roughly ten to forty percent depending on the demographic model.⁴ The Roman economy ran on coerced labor at a scale the modern world has difficulty fully comprehending. For the senatorial and equestrian elite, together comprising less than one percent of the empire’s population of roughly sixty million,⁵ this meant genuine exemption from physical and commercial labor. Their income came from land worked by slaves. Their time, in theory, belonged entirely to themselves.

What did they do with it? The ideal, among educated Romans, was something close to what we might call a Renaissance life: philosophy, rhetoric, literature, military service, cultivation of the self and the civic order. Cicero wrote his greatest works in enforced political retirement. Marcus Aurelius composed his Meditations on campaign. Seneca produced some of the most enduring thinking about the right use of time that any human being has ever produced. These were men who had the gift of leisure and used it to think carefully about what leisure was actually for.

But the Stoics (Seneca, Epictetus, Marcus Aurelius) were not describing Roman society as it was. They were writing in reaction to what Roman society was becoming. Their philosophy was an attempt to answer a question that slavery had forced into the open: how do you live well when external necessity is removed? How do you build a meaningful life when the world no longer needs you to show up?

The answer most Romans gave was not the Stoic answer. It was spectacle and dependency. By the mid-first century CE under Claudius, the Roman calendar included more than 150 days designated as public holidays, 93 of them devoted to publicly funded games.⁶ Gladiatorial games, chariot racing, public executions, elaborate theatrical performances. These were not merely entertainment. They were the primary occupation of enormous portions of the urban population. Combined with the grain dole (free monthly distributions to approximately 200,000 adult male citizens under Augustus⁷), you had something that looks, from a distance of two millennia, remarkably like a universal basic income paired with endlessly available entertainment. Bread and circuses. The phrase has survived because the pattern it describes is not Roman. It is human.

And the working class of Rome (the free artisans, the farmers, the craftsmen) fared worst of all. Slavery did not liberate them. It undercut them. Why hire a free carpenter when a slave carpenter costs nothing beyond upkeep? The Roman working class was not exempted from labor by the existence of slavery. They were economically marginalized and culturally humiliated by it. Their work became associated with servitude. Their dignity eroded alongside their wages. They did not gain leisure. They gained precarity.

III. The Meaning Problem

Every previous wave of automation primarily displaced physical labor or routine cognitive labor.⁸ What came next was always more complex, more creative, more “human.” The assumption embedded in two centuries of economic thinking about technology is that we would always be needed at the top of the cognitive pyramid. That assumption is now, for the first time, genuinely in question.

But the problem is not only economic. It is also psychological, and it runs much closer to the bone.

Work is not merely a source of income. It is, for most people, the primary mechanism through which a meaningful life is assembled. It provides identity: the first thing humans say about themselves is what they do. It provides competence: the felt sense of getting better at something, of being genuinely good at a thing that matters. It provides stakes: the knowledge that something real depends on whether you show up. It provides structure, progress, belonging, and rhythm. It bundles an enormous amount of what makes a life feel worth living into a single daily activity that most people access without having to think about it.

The psychologist Mihaly Csikszentmihalyi spent decades studying the states of deep absorption and satisfaction that humans call flow, the condition in which challenge and skill meet perfectly and time disappears. In a study using experience sampling with adult workers, he and Judith LeFevre found that flow states occurred far more frequently during work than during leisure, though participants paradoxically reported wishing to be at leisure rather than working.⁹ We believe we want more free time. Unstructured free time, in practice, often produces less meaning than the work we were trying to escape.

This points to something even more fundamental. There is something almost philosophical about the act of applying yourself to resistant reality, where the world pushes back and you push through anyway. A surgeon cannot fake a surgery. A carpenter cannot fake a joint that holds. A farmer cannot negotiate with a drought. That resistance is not incidental to the meaning. It is the meaning. The struggle is what makes the competence real, the accomplishment earned rather than granted.

AI threatens to remove exactly that friction. And if it does so too completely, too quickly, it may inadvertently remove the primary mechanism through which most people construct a sense that their life is worth living. This is not a small thing to remove. We have not built a replacement.

IV. What Is Already Happening

It would be more comfortable to treat all of this as a thought experiment, a problem for the next generation to navigate. But the displacement is already underway. And so is the resistance.

In 2025, companies in the United States cited artificial intelligence in announcing approximately 55,000 job cuts, more than twelve times the number attributed to AI just two years earlier.¹⁰ Microsoft confirmed that roughly thirty percent of its code was being written by AI before laying off approximately 6,000 employees, the majority of them programmers. Klarna reduced its workforce from 7,000 to 3,000 after deploying an AI assistant that handled the equivalent workload of 700 full-time customer service agents. Amazon announced 30,000 corporate job cuts across 2025 and early 2026, with its CEO acknowledging plans to use AI to “reduce our total corporate workforce as we get efficiency gains.” These are not projections. They are earnings calls.

The resistance is equally real, and arguably more revealing. In October 2024, the International Longshoremen’s Association shut down every major East Coast and Gulf Coast port in a strike whose central demand was not wages but a ban on the automation of cranes, gates, and container-moving trucks. The Writers Guild of America struck for 148 days in 2023, driven in large part by the threat of AI-generated scripts undercutting writers’ livelihoods and credit. SAG-AFTRA followed, fighting for consent protections before studios could create digital replicas of performers. In Las Vegas, hospitality workers negotiated contracts requiring six months’ advance notice before any AI deployment.¹¹

What is striking about these disputes is not just their scale. It is what the workers are demanding. They are not asking for retraining programs or transition assistance, the standard technocratic response to displacement. They are asking not to be replaced. The distinction matters. It suggests that what is at stake, for the people closest to the transition, is not simply income. It is the thing the previous section describes: identity, competence, the sense that your contribution matters. The resistance is a meaning problem dressed in the language of labor negotiation.

The freelance market offers a quieter but equally telling signal. A peer-reviewed study published in Organization Science found that after ChatGPT’s release, freelancers on major platforms in AI-exposed occupations experienced measurable declines in both contracts and earnings, and the effects were most pronounced among high-skilled, high-priced freelancers, contradicting the assumption that experience would provide protection.¹² And the shift is not only on the worker side: analysis of firm-level spending data found that more than half the businesses using freelance platforms in 2022 had stopped entirely by 2025, redirecting that spend toward AI providers instead.

This points to something the economic data confirms at a larger scale. A study published in Science by researchers at OpenAI and the University of Pennsylvania found that approximately eighty percent of the U.S. workforce could have at least ten percent of their work tasks affected by large language models, with higher-income occupations facing the greatest exposure.¹³ This reverses the pattern of every previous automation wave, which displaced labor from the bottom of the skill distribution upward. AI is climbing the cognitive pyramid from the top.

And the people closest to this transition know it. A 2025 Pew Research Center survey found that fifty-two percent of American workers are worried about AI’s future impact in the workplace. An EY survey focused specifically on agentic AI (systems that can plan, reason, and execute multi-step tasks autonomously) found that while eighty-four percent of employees were eager to work with such systems, fifty-six percent simultaneously worried about their job security, and fifty-one percent worried that agentic AI would make their jobs obsolete.¹⁴ The eagerness and the dread are not contradictory. They are the precise emotional signature of a transition that people can see coming but feel powerless to stop.

I see this dynamic in my own work, every week. The executives and senior leaders I work with want agentic AI solutions. They see the efficiency gains and the competitive pressure and they want to move fast. The working-level employees whose tasks those systems would absorb see something else entirely. They push back. And they are right to. They are not resisting out of ignorance or stubbornness. They are the people who can see most clearly what is coming, because it is coming for what they do every day.

But it would be dishonest to frame this as a simple story of replacement. For most people right now, including most of my customers, the immediate question is not whether they will lose their jobs. It is what their jobs are becoming. Can a financial analyst use AI to process in minutes what used to take days, and spend the freed time on judgment calls that actually matter? Can a marketing team produce twice the output at higher quality, not by replacing people but by changing what “doing the work” means? Can a developer shift from writing boilerplate to designing systems, because AI handles the implementation? In many cases, the honest answer is yes. The people I work with are not all being displaced. Many of them are being elevated, doing more, doing it faster, and doing more valuable work than they could before. That is real, and it matters. The augmentation story is not a fantasy. It is happening alongside the displacement story, often in the same company, sometimes in the same team. The question is how long it lasts: whether augmentation is a stable new arrangement, or a transitional phase on the way to something more complete. The analyst whose AI handles the routine work may thrive for a decade. But the analyst whose AI handles the routine work and the judgment calls is something else entirely. The line between tool and replacement is not fixed. It moves. And it has been moving in one direction.

V. The Class Divide Worth Examining

The popular narrative about AI disruption is democratic in its distribution of anxiety. Everyone is disrupted. White collar, blue collar, all of us together, navigating an uncertain future. There is something comforting in this framing. It suggests solidarity. It implies shared stakes.

It may not hold up under scrutiny. The experience of this transition is already radically uneven, and a few encounters have stayed with me.

I once asked a wealthy acquaintance what he would do when fully autonomous vehicles arrive. He laughed. He already has a driver. He sleeps in the back seat on his way to work. The question was meaningless to him. The post-work world, for people at his level of wealth, is not a future scenario. It is Tuesday.

At the other end, a college student recently told me that his friends are struggling, not with the familiar question of whether they are studying the right major to land a good job, but with a newer and more unsettling one: will there be anything for them to do at all? The question has shifted from optimization to existence. That is a different kind of anxiety entirely.

And between those poles are the people whose relationship to work is already broken, people doing jobs they hate, jobs that drain them, jobs they endure for the paycheck and nothing else. For them, the promise that AI will take over the work might sound like liberation. But liberation into what? If work is the primary mechanism through which people construct meaning (and most of the evidence suggests it is), then being freed from work you hate does not automatically deliver you into a life you love. It delivers you into a vacuum. The meaning problem does not spare the people who hated the work. It may reach them last, but it will reach them.

We do not have to speculate about this. We already have millions of people living without work, and what they experience is revealing. Research on retirement shows that the transition can go either way: people who were dissatisfied with their jobs often find their sense of purpose increases after leaving, while those whose identities were tightly bound to their careers face elevated rates of depression, cognitive decline, and loss of meaning.¹⁵ The determining factor is not retirement itself. It is whether someone had built a structure for meaning outside of work before the work was removed. The financial independence movement offers a similar lesson at a smaller scale. Forums are filled with people who achieved the dream of never needing to work again, only to find themselves adrift within months, not because they missed the work but because they had never developed anything to replace what the work had quietly provided. If retirement is a preview of the post-work world, its message is not that the transition will be catastrophic. It is that the transition will be catastrophic for the people who were not prepared, and right now, almost no one is being prepared.

What seems more likely, based on both the historical pattern and the current economic trajectory, is a bifurcation that maps uncomfortably onto the Roman model. A small class of people who own, direct, and leverage AI may become extraordinarily powerful. A large class of people whose skills are replaced by AI may face something close to the Roman working-class fate: economically marginal, culturally adrift, and dependent on whatever the state or the platforms decide to provide. Between them, a thin middle of people who can collaborate with AI in ways that amplify rather than replace them, but this window would require continuous adaptation and may not remain open long.

It is worth asking who is most exposed. Physical trades (plumbing, electrical work, construction) appear safer for now, because they require embodied presence and manual dexterity that AI cannot yet reach. The educated middle class may be the new free Roman artisan: the writers, analysts, paralegals, junior developers, customer service workers, graphic designers, entry-level accountants. These are roles where AI is already capable of doing in seconds what used to take hours. The extent and speed of this displacement is an open question, but it is a question worth taking seriously rather than assuming away.

And the deepest damage may not be economic. What really hollowed out Rome’s working class was not just the loss of income. It was the loss of narrative importance. They had been the backbone: the farmers, soldiers, and craftsmen who built the Republic. Slavery did not merely take their jobs. It took their place in the story of civilization. The danger worth examining is not just unemployment. It is irrelevance, the feeling that the future is being built without you, for you at best, around you at worst.

VI. The Asymptote

Here is what I find most striking, as someone who builds these systems for a living: the question of what humans are for, once the work is done, does not seem to be getting the sustained attention it deserves from the people building the systems that make it urgent.

This is not an accusation of bad faith. The engineers and researchers building AI are, in my experience, largely thoughtful people who care about the consequences of what they are making. The problem may be structural. The logic of the industry (move fast, optimize, scale, ship) is not naturally suited for sitting with an irresolvable human question. And so the question tends to get deferred, reframed as an economic problem, or quietly assumed away. The answer will emerge, we are told. People will find new things to do. They always have.

They have. Until now, they have. And the counterargument deserves to be taken seriously: every previous doomsday prediction about technology and employment has been wrong. The loom did not end weaving. The ATM did not end banking. The spreadsheet did not end accounting. New tools create new possibilities, new industries, new forms of work that no one anticipated beforehand. Perhaps AI will do the same. Perhaps it will generate roles we cannot yet imagine, in fields that do not yet exist, requiring skills we have not yet named. This is not a foolish argument. It has two centuries of evidence behind it.

There is also a nearer-term version of this argument, and it is the one I see playing out with my own customers: AI does not replace the worker. It transforms the work. The analyst becomes a strategist. The developer becomes an architect. The customer service agent becomes a relationship manager. People do more, do it faster, and do more valuable things than they could before. This is genuinely happening, and in the short term it may be the dominant story. But even the augmentation scenario changes what work means. If AI handles the analysis and you review its output, the felt experience of competence shifts. You are no longer the one who wrestled the data into insight. You are the one who checked whether the machine got it right. The skill changes from doing to supervising, from creating to curating. That is not nothing. Oversight and judgment are real skills. But they are different skills, and they provide a different kind of meaning. The question the meaning problem poses is not only what happens when work disappears. It is what happens when work is hollowed out, when the parts that gave it weight are delegated to a system that does them faster and, increasingly, better.

But the counterargument has a structural weakness. As this essay has already argued, every previous automation wave displaced labor at the bottom of the cognitive stack and created new demand at the top. AI is the first technology that can follow humans up the stack. It can, at least partially, do the creative, the analytical, the relational, the written, the reasoned. The escape hatch has always been “move up.” What happens when there is no up left to move to? The argument that new jobs will simply emerge is not wrong. But it is a bet, not a certainty. And we are making that bet on behalf of billions of people without asking them whether they accept the odds.

The asymptote is the right metaphor for this. Science and technology have always operated on the implicit promise that they are converging on the answers, that with enough data, enough compute, enough optimization, the important questions will eventually yield. And on many questions, this has been spectacularly true. We have cured diseases, mapped genomes, put instruments at the edge of the solar system. The progress is real and it is genuinely extraordinary.

But the curve never touches the line. The closer technology gets to the questions that matter most about human life (what is consciousness, what makes suffering meaningful, what a life should be for, what we owe each other), the more clearly it becomes apparent that these questions are not the kind that yield to optimization. They are not data problems. They are not compute problems. They are the permanent remainder of everything that matters, left over after the algorithm has taken everything it can reach.

VII. The Question That Stays Open

I am raising children who will inherit whatever we build in the next decade. That fact sharpens the question considerably. I can teach my kids to code, to prompt, to leverage AI tools. I can prepare them for the technical landscape they will enter. But the more important thing I can give them, and the harder thing to give, is a framework for deriving meaning from a world that may not need them to work the way I have worked, the way my parents worked, the way every generation before them worked.

And if that is true for my children, it is true for everyone’s. The retirement research in the previous section points to something unsettling: if the post-work world arrives broadly, it will be as if an entire generation retires before ever starting a career. And the institution most responsible for preparing people for that transition, education, is not designed for it.

The modern school system, from kindergarten through graduate school, is built around producing workers. Choose a major. Build a resume. Acquire marketable skills. The entire structure assumes that the point of learning is employment. If employment becomes optional, or unavailable, for large portions of the population, then education faces a crisis of purpose that mirrors the one facing individuals. Does schooling become optional? Does it transform into something closer to what the liberal arts once promised: not job training but the cultivation of a person capable of thinking, creating, and living well? Or does it simply hollow out, the way Roman civic institutions hollowed out once the practical reasons for them disappeared? We do not have an answer to this yet. But the question is no longer hypothetical. It belongs to the students sitting in classrooms right now.

I do not have a clean answer to any of this. I want to be honest about that. Anyone offering a clean answer to this question is selling something.

What I do believe is this: the people who will navigate this transition best are those who have built, deliberately and before the necessity is removed, a relationship with meaning that does not depend entirely on external usefulness. People who can find richness in presence, in craft practiced for its own sake, in genuine relationships, in the kind of deep attention that productivity culture has always told us is a luxury. The Stoic insight (that a good life requires chosen discipline, not imposed necessity) turns out to be the most practically urgent philosophical idea of the twenty-first century.

The Stoics are not alone in this. Contemplative traditions across cultures have spent millennia thinking about purpose beyond productivity, and their thinking deserves attention now. Buddhist practice treats right livelihood as one component of a meaningful life, not its totality, and cultivates presence and attention as disciplines in themselves. Christian monasticism built entire communities around structured meaning without market-facing work. The Benedictine principle ora et labora (pray and work) defined work as service and contemplation, not output. The Japanese concept of ikigai, a reason for being, explicitly includes but is not limited to what you are paid for. These are not quaint relics. They are functioning frameworks, tested across centuries and civilizations, for answering exactly the question this essay is asking. The fact that billions of people already live within them suggests that the post-work meaning problem is not unsolvable. But it does require the kind of deliberate cultural infrastructure that our current economic system has not been designed to provide.

But I also believe that individual virtue is not sufficient. Modern historians have largely moved past the idea that Rome declined because Romans became morally weak; the causes were structural, economic, military, and demographic.¹⁶ But there is something in the Roman story that I find hard to set aside: once slavery removed the need for most people’s contribution, Roman society had no structure for answering the question of what those people were for. Bread and circuses was not malicious. It was the path of least resistance. It is always the path of least resistance. And unless we build something better, consciously and structurally, it will be our path too.

The window to shape that is open right now. It will not remain open indefinitely. The question of what humans are for, in a world where AI does most of the work, is not confined to any single discipline. It is the central design challenge of our moment.

But here is one thing I keep coming back to, one thing that does not feel like an open question: AI cannot be accountable. It can generate, optimize, execute, and even reason, but it cannot be the one who answers for what it does. It cannot bear the weight of a decision that affects someone’s life and say I chose this, and I own what follows. Accountability requires a person: someone who can be held responsible, who accepts consequences, who stands behind the judgment call when things go wrong. If there is a role that remains irreducibly human, it may be this: not doing the work, but being answerable for it. The question is whether we will accept that role deliberately, or let it dissolve by default, the way Rome let citizenship dissolve into spectatorship.

I ask these questions with my children in mind, with my team and my customers in mind, with the society we are all building in mind. But if I am honest, I am asking them primarily for myself, as I try to figure out my own next steps, and try my best to serve the people around me while the ground shifts under all of us.

Notes