Editor's Note: Whether AI will cause mass unemployment is one of the most common technological anxieties today.

This article argues that this 'AI employment apocalypse narrative' is not new. It essentially follows the old logic of 'fixed total amount of work,' suggesting there is a fixed amount of work in society, and the more AI does, the less humans can do.

The author counters that history hasn't unfolded this way. Tractors reduced the number of farmers but gave rise to manufacturing, services, and software industries; electricity replaced older power sources but reorganized factories and consumer goods industries; Excel didn't eliminate finance jobs but instead spawned more financial analyst positions. Technology does eliminate certain tasks and jobs, but the larger result has often been the creation of new demands and new jobs.

Applied to AI, what truly deserves attention is not 'will humans be completely replaced' but 'which jobs will be redefined by AI.' AI will compress some repetitive labor, but it will also make roles like software engineers, product managers, and systems designers more efficient and important. In other words, AI is more like a productivity tool than a mere job killer.

The core judgment of this article is: AI will not bring the economy to a halt; it will only force a reallocation in the labor market. Doomsayers only see the jobs being replaced but ignore how technological revolutions have consistently created new industries, new professions, and new spaces for growth.

The original text follows:

The panic over a 'permanent underclass' touted by AI pessimists is not a compelling narrative. It's not even a new story. It's merely a repackaged version of the 'lump of labor fallacy.'

The 'lump of labor fallacy' refers to the view that the total amount of work needed in society is fixed. It assumes a zero-sum competition between existing workers and any agent that might perform similar work—whether that's other workers, machines, or this time, AI. If the total amount of useful work is fixed, then the more AI does, the less there must be for humans to do.

The problem with this premise is that it contradicts our basic understanding of people, markets, and economies. Human desires and needs have never been fixed. About a century ago, Keynes famously predicted that automation would lead to a 15-hour workweek, but history proved Keynes wrong. He was right about one thing: automation did create a 'labor surplus'; but instead of lying back to enjoy leisure, we consistently found new, different productive activities to fill our time.

Of course, AI will definitely eliminate some tasks and compress some job roles—and there is already some evidence suggesting this change may be underway. The shape of the labor market will change, just as it has with every transformative technology unleashed. But claiming that AI will cause economy-wide, permanent unemployment is unhelpful marketing rhetoric, poor economic judgment, and a misreading of history. On the contrary, productivity gains should increase the demand for labor because labor itself becomes more valuable.

Here is our argument.

Humans, Checkmated?

We agree with the doomsayers' assessment—frankly, anyone with their eyes open can see it—the price of cognition is collapsing. AI is becoming increasingly adept at handling tasks that, until very recently, were considered the exclusive domain of the human brain.

The doomsayer's logic is: 'If AI can think for us, then humanity's 'moat' disappears, and our terminal value goes to zero.' Checkmate, humans. As if we have completed all the thinking we need and want to do; and now, since AI will shoulder an increasing share of the cognitive load, humans can only slide towards obsolescence and uselessness.

But the problem is: both historical precedent and intuition tell us that when the cost of a powerful input falls, the economy doesn't politely stand still. Costs fall, quality improves, speed increases, new products become viable, and demand curves shift outward. Jevons' Paradox still holds here. When fossil fuels initially made energy cheap and abundant, we didn't just put whalers and lumberjacks out of work; we also invented plastic.

Contrary to the doomsayers, we have every reason to expect a similar effect from AI. Since AI will shoulder an increasing share of the cognitive load, humans can instead free up their hands to explore frontiers more ambitious than ever before.

If historical experience remains instructive, then we can expect technological change to make the entire economic pie larger.

Every once-dominant economic sector has eventually given way to a larger successor; which in turn further propelled the overall economy to become even larger.

Today, the tech industry's scale already exceeds that once achieved by finance, railroads, or the industrial sector, yet its share of the entire economy or the entire market is still smaller. Productivity gains are far from a zero-sum game; they are an immensely amplified positive-sum force. After we handed over so much labor to machines, the ultimate result was not a shrinking economy and labor market, but ones that became larger, more diverse, and more complex.

Doomsayers want you to ignore the history of innovation, focus only on the single frame of 'cognition costs are collapsing,' and present it as the entire movie. They see task replacement and stop there.

'Our cognitive output will increase tenfold, but we won't think about more things as a result; we'll pat our stomachs and go to lunch early; so will everyone else.' This idea not only reveals a significant lack of imagination but also a failure to observe basic reality. Doomsayers call this 'realism,' but this is not what has truly happened in history—never.

The Failure of Luddism

Let's look at what actually happens when a great leap in productivity sweeps through the economy.

Agriculture

Before the widespread adoption of agricultural mechanization in the early 20th century, about one-third of the US working population was engaged in agriculture. By 2017, this proportion had fallen to about 2%.

If automation caused permanent unemployment, tractors should have already destroyed the labor market. Yet the opposite happened: agricultural output nearly tripled, supporting massive population growth; and those workers who left the farms did not become permanently unemployed but flowed into previously unimaginable new industries—factories, shops, offices, hospitals, laboratories, and eventually into the service and software sectors.

So, of course, it can be said that technology did disrupt the career prospects of the average farmworker; but it was precisely in this process that it unleashed a global surplus of labor and resources and ushered in a whole new economic system.

Electrification

The story of electricity is similar.

Electrification wasn't just about swapping one power source for another. It replaced central line shafts and belts with individual electric motors, forcing factories to reorganize around entirely new workflows and creating entirely new categories of consumer and industrial goods.

This is a typical feature of how technological revolutions unfold in different phases. Carlota Perez documented this process in 'Technological Revolutions and Financial Capital': a phase of massive investment and intense interest from financial capital appears early on, the cost of durable goods plummets, followed by a generational growth cycle for durable goods manufacturers.

Electricity also took a considerable time to truly unleash its productivity magic. At the turn of the 20th century, only 5% of US factories used electricity to drive machinery, and less than 10% of homes were electrified.

By 1930, electricity supplied nearly 80% of the power for US manufacturing, and labor productivity growth rates doubled in the following decades.

It did not destroy the demand for labor. On the contrary, higher productivity meant more manufacturing activity, more salespeople, more credit issuance, and more commercial activity. Not to mention the second-order effects of labor-saving devices like washing machines and cars—they both freed more people from previously inefficient labor, allowing them to enter higher-value activities that were impossible before.

As car prices fell, both car production and employment saw explosive growth. This is the result of a true general-purpose technology: it reorganizes the economy and expands the boundaries of 'useful work.'

We have seen this time and again. Did VisiCalc and Excel spell doom for bookkeepers? Clearly not. Computational technology that drastically improved efficiency actually led to an expansion in bookkeeping positions and created the entire FP&A (Financial Planning & Analysis) industry.

We lost roughly 1 million 'bookkeeper' jobs but gained about 1.5 million 'financial analyst' jobs.

Those New Service Jobs

Of course, task substitution doesn't always lead to job growth in adjacent areas of the economy. Sometimes, the surplus released by productivity manifests as entirely new employment growth in a completely unrelated sector.

But what if AI means some people become extremely wealthy while others are left behind?

At the very least, those extremely wealthy people will have to spend their money somewhere, creating brand-new services from scratch. This is precisely what happened in the past:

Massive productivity gains and the subsequent wealth creation they brought forth gave rise to many entirely new job types. Without rising incomes and increased labor availability, these jobs might never have truly emerged—even if they were technically feasible before the 1990s. Regardless of how one views service jobs for the wealthy, the end result was that everyone became better off: higher demand drove a significant rise in median wages, thereby creating more 'wealthy' people.

Stripe's internal economist, Ernie Tedeschi, provides a very interesting 'synthetic case': how a job can be impacted, transformed, and reshaped by technology—the travel agent.

Did technology reduce the demand for travel agents? Yes, certainly:

Today, travel agency employment is about half what it was at the turn of the century, almost certainly due to technology.

So, does this mean technology is a job killer? The answer is still no. Because travel agents didn't become permanently unemployed. They found work elsewhere in the economy; overall, the age-adjusted employment-to-population ratio today is roughly the same as in 2000.

Meanwhile, for those travel agents who remain in this now technology-empowered industry, productivity gains mean their wages are higher than before:

'In the golden era of 2000, the average weekly salary for a travel agent was equivalent to 87% of the overall average weekly salary. By 2025, this ratio had reached 99%, meaning travel agent wages grew faster than the overall private sector during this period.'

So, even in this case where technology did severely impact the scale of travel agent employment, overall employment among the working-age population remained as stable as before, and the travel agents who remained are actually better off than ever.

Augmentation > Substitution, and the Jobs That Don't Exist Yet

This last point is crucial and again shows that doomsayers only tell a very small part of the whole story.

For some jobs, AI is indeed an existential threat. That's true. But for other jobs, AI is a force multiplier that will make these roles more valuable. Next to every job at risk of AI substitution, there are other jobs that might benefit from AI:

Goldman Sachs's estimated 'AI substitution' effect has been offset, and even overshadowed, by the 'AI augmentation' effect. Management also seems to be focusing more on 'augmentation' than 'substitution'—a point equally worth noting:

To date, mentions of 'AI as an augmentation tool' on corporate earnings calls are about 8 times more frequent than mentions of 'AI as a substitution tool.'

Although Goldman Sachs didn't even include software engineers on its 'augmentation' list, software engineers are likely the best example of an AI-augmented role. AI is a force multiplier for writing code. Not only are Git commits soaring, but the number of new applications and new companies is increasing, and the demand for software engineers appears to be on an upward inflection:

Whether measured by the number of job openings or their share of the overall job market, software development roles have been increasing since early 2025.

Is this because of AI? Honestly, it might still be too early to conclude definitively. But there's no doubt that AI enhances software engineering capabilities, not to mention that AI has become a top concern for every company and every executive.

When everyone is trying to figure out how to integrate AI into their business, companies naturally have reason to hire heavily to achieve this goal. This makes certain employees more valuable, not less:

Jobs with higher AI exposure seem to be driving above-trend wage growth; this is particularly evident in systems design roles.

These gains may still be relatively concentrated for now, but it's still very, very early. As professional capabilities diffuse, opportunities will spread accordingly. In any case, this is not the data doomsayers want you to see.

Meanwhile, according to Lenny Rachitsky—author of Lenny's Newsletter and a key insider in the tech community—open product manager roles continue to rise, having recovered from the interest rate-driven collapse and reached their most abundant level since 2022:

The simultaneous growth in hiring for software engineers and product managers is a concise example of why the 'lump of labor fallacy' is wrong. If AI substituted for thinking in a 1:1 manner, you might reasonably expect: 'product managers need fewer engineers,' or conversely, 'engineers need fewer product managers.' But that's not the reality we see. What we see is that demand for both types of roles continues to rebound because what truly matters is: people can accomplish more work.

This is why the doomsayers' failure is, at its core, a failure of imagination. They only stare at the tasks being automated away but ignore a new demand frontier—one that will create jobs we haven't even imagined yet:

Most new jobs created since 1940 didn't even exist in 1940. By 2000, it was easy to imagine many travel agents losing their jobs, but it was probably hard to imagine then that a mid-market tech services industry built around 'cloud migration' would emerge—after all, the true rise of cloud computing was still over a decade away.