Since ChatGPT’s paradigm-shifting debut in November 2022, writers and thinkers of all stripes have touted the virtues and lamented the dangers of A.I. Such dangers include A.I.’s disastrous effect on education and on creativity, its potential to enable ordinary citizens to make weapons of mass destruction or to flood the world with deepfakes, its supposed tendency to reinforce stereotypes, its willingness to steal our intellectual property or its capacity to kill us off entirely.

Yet one danger of A.I. looms above all others — and we’re completely unprepared for it: the coming wave of A.I.-driven unemployment likely to hit within the next few years.

If you disagree — and judging by the rosy forecasts I frequently hear promulgated by our ostensible “experts,” many disagree — allow me to explain. I am an attorney educated in our elite Ivy League institutions in the years prior to their full-scale devolution into anti-intellectual halls of cracked mirrors amplifying and reinforcing the shallow ideological prejudices of insular elites. I have 25 years of experience in complex commercial litigation often involving abstruse legal issues and requiring a certain level of analysis, creativity, strategic thinking, persuasive writing and hard-won knowledge. That is not intended as an ad for my services. It is intended, rather, as a set-up for the rather startling fact that as a direct consequence of A.I., my services and the similar services of many white-collar professionals like myself may soon be substantially less necessary.

Our staunchly capitalist citizens and leaders are not going to transition seamlessly from mass employment to … living on the dole.

My job has become significantly easier with every year since A.I.’s public launch in November 2022. Document review has become infinitely less time-consuming, with A.I. able to help me zero in on the comparatively few key documents likely to contain useful information or key admissions by the opposing side.

Legal research that once consumed many hours, with much time wasted on dead-end trails, now often consumes a small fraction of that time, as the right prompt summons up the relevant statutes or cases I need to get me going in the right direction or even hand me the ultimate answer I need. Complex regulatory schemes become easy to wade through with some A.I. handholding. Searches for good forms for specific court filings now proceed much more efficiently. While I haven’t yet used A.I. for brief-writing, I know many lawyers have, at least to give themselves a rough draft to work from.

With all this A.I. facilitation, many tasks that used to take five hours now take one or two. The pace of progress since 2022 has been incredible. When A.I. first came out, it was near-useless at work. Even when ChatGPT-4 was the going model throughout much of this year, it was constantly plagued by hallucinations, where it would make up quotes from fake cases or utterly misstate their doctrines. Many of these issues have been dramatically resolved with the unveiling of ChatGPT-5 in early August (OpenAI reports a 78 percent drop in the hallucination rate from prior versions). And I am not even using a premium version of ChatGPT and am far from any sort of A.I.-tech whiz.

Now, consider that for decades, large, prestigious law firms have been hiring incoming “classes” of dozens upon dozens of recent law school graduates. These small lawyer armies are — or were — necessary for the grunt-work of sorting through terabytes of data in document-intensive cases big firms routinely handle. Such starting lawyers would also spend many hours doing often inefficient but necessary research tying up every conceivable loose end. In the process, these junior attorneys, often working late nights for weeks on end, would gain invaluable knowledge and skills atop which they could build the rest of their careers.

And here we come to the nub of the matter: if, after some three years of living with A.I., tasks that used to take me five hours now often take less than half that time, what do you think will be A.I.’s impact on the basic but labor-intensive grunt-work for which starting lawyers are typically enlisted? And how long do you think it will take these large, bottom-line oriented law factories to realize that what used to be the job of 100 inexperienced incoming lawyers can now be done by less than half that number?

Now, project this out another two or three or five years into the future, when, if the current pace of progress is any guide, the capacities of A.I. will be that much greater. How many starting lawyers do you think it will take in five years to do today’s 100-lawyer job? Maybe 20 … or even 10. This is why I believe the employment apocalypse is no more than — and probably less than — five years away.

The looming threat extends far beyond the legal industry, of course. Pick any white-collar industry: accounting, journalism, finance, software engineering, marketing, education, academic research, customer service, communications and so on. We can expect the same sort of employment-market carnage in all of these domains. Those just starting out in their careers and with the least to bring to the table are going to be hit the hardest, but even those higher up in the hierarchy will not be immune.

Early data supports my anecdotal claims that A.I. increases worker efficiency. A 2023 study published in Science found that in completing designed writing tasks similar to components of their jobs, workers saw a 40 percent improvement in speed and an 18 percent improvement in quality.  A 2024 Future of Professionals report from Thomson Reuters found lawyers believed that within five years, A.I. would be saving them 12 hours of work per week. A June 2025 PwC report found the industries most exposed to A.I. — such as financial services and software engineering — had already seen a threefold increase in revenue per employee.

Ironically, credentialed “knowledge” workers, those who previously thought themselves least susceptible to economic shakeups, are first in line to be hit by the speeding A.I. tidal wave. But it is only a matter of time for other industries where skilled or unskilled manual labor predominates.

Driving, an industry where so many immigrant men get their start in this country, is already being automated away (though an effort is underway to jam on the brakes to some limited extent). But take more “skilled” blue-collar jobs, such as plumbing or auto repair. How long do you think it will be before A.I. — especially when plenipotent A.G.I. comes on the scene, which now, according to the experts, seems to be a matter of a few years at most — engineers machines to discharge such tasks?

Or take another “elite” job, the realm of medicine, which largely involves knowledge work but sometimes also has skilled manual components in the form of (increasingly rare) physical exams or surgery. Especially now that it’s cut down on hallucination, A.I. is already superior to many doctors (and certainly those forced by insurers to spend no more than 15 minutes per patient) at sorting through research to figure out what your various symptoms may amount to or which of your prescriptions or supplements might be causing that nasty rash. As such, putting most doctors out of a job is on the horizon, with only the far smaller physical part of the job likely taking a bit longer to displace.

What jobs will remain after the employment apocalypse hits? Very high-end jobs, perhaps, those at the tops of our hierarchies. Jobs that involve manufacturing handmade, artisanal, eclectic curios of the sort sold on a platform like Etsy. Jobs that have substantial creative components. ChatGPT’s creative writing, for example, is still pretty rough, with its sentences replete with clichés or strained, clunky similes and metaphors. But, again, quantum leaps in this area are probably a matter of (not too much) time.

Unless ChatGPT can “learn” to taste and smell, replacing creative chefs seems a steeper hill to climb. Jobs protected by professional associations. Attorney bar associations, medical licensing boards, actuarial societies and other similar organizations will surely try to keep jobs in human hands. But again, these will likely be the jobs at the top of the food chain.

Jobs that have some ineffable and irreplaceable human component. For example, because there is evidence for physiological synchrony (in measures such as heart rate and heart rate variability) as well as support for the role of mirror neurons during psychotherapy sessions, I can imagine silicon-based life will never fully replace the magic of human-to-human interaction. A similar phenomenon could occur in a job like nursing, an area OpenAI’s Sam Altman has speculated would be more immune to the ravages of A.I.

In all events, however, within just a few short years, our diverse present-day workplace is going to be decimated, with only a handful of highly qualified experts or A.I.-supervisors doing what is now being done by much of our adult population. Before turning to the implications, I want to pause briefly to address a few objections.

This same species of doomsaying was common enough during prior waves of industrialization and technological displacement. Those predictions proved wrong, as new technologies ushered in new, unanticipated career paths. A.I., this argument would go, will do the same, as there will be jobs prompting and supervising A.I. and other jobs we cannot yet even envision.

My response is two-fold. First, what prior rounds of technological innovation did is automate away much skilled and unskilled manual labor and create demand for knowledge workers able to engineer, operate, and service the machines. The result has been an ever-greater push for us to educate ourselves to meet the demands of a high-tech marketplace. Initially, bachelor’s degrees were sufficient; more recently, post-grad degrees have become a more commonplace expectation. It is apparent, however, we are at the limit of our educable capacities. With falling math scores, a cratering capacity to read difficult texts and even declining IQ scores, we do not seem capable of absorbing yet more education, not to mention the absurdity of having to spend yet more years incurring debt before finally joining the labor market in what would be our late 20s.

But second — and this response is even more devastating — the whole devilish magic of A.I. is that it replaces knowledge workers first and foremost. It will only get better at doing so in the coming years. Like chess grandmasters, who ultimately conceded the battle to their silicon-based adversaries, we will be unable to outpace A.I., unable to identify or create new jobs it will be incapable of doing better. The whole point, especially after A.G.I. comes on the scene, is that this is a general purpose tool capable of adapting itself to nearly every intellectual — and eventually physical — challenge we can dream up.

Another objection to my thesis is that, so far, we have seen no evidence of job loss. First, that’s not actually true. A recent Stanford study noted a 13 percent decline in employment among early-career workers in the tech sector, with the most severe declines being — just as we would expect if A.I. is the culprit — among entry-level employees. As a recent Forbes article explains:

The report says it’s uncovered “substantial” declines in employment, especially for workers ages 22 to 25. This dovetails with mounting evidence from investment banks and surveys of layoff announcements, as Goldman Sachs has calculated a shrinking premium from a college degree…. Bank of America Global Research, meanwhile, has noted that since 2022, the unemployment rate for recent graduates has started to exceed the overall unemployment rate for the first time in recent memory.

Due to the particular fact that recent college graduates are having difficulty finding employment, I would speculate at least some portion of the slowdown in hiring reflected in the most recent U.S. jobs report is due not to President Trump’s tariffs, but rather, to A.I. starting to take its toll.

But this is just the beginning. A.I.’s impact on productivity is a very recent phenomenon. Employers need time to process what is going on. Most, with reputations at stake, are going to be hesitant to overreact and get themselves featured in the news for A.I.-driven mass layoffs. But after just one bold move on the part of one big employer gets the ball rolling, a bunch of copycats will follow in stride, whereupon the avalanche will hit. And, as I said, it’s coming a lot sooner than you think.

Another possible objection might be that if companies and governments see the massacre unfolding, they will press pause on A.I. But they won’t because they can’t. For companies, A.I. increases their productivity and profit margins. This is especially true of companies that create A.I. engines. They will do all they can to stay ahead of their competitors. The same prisoner’s dilemma affects governments. If the U.S. halts A.I. progress, China or some other competitor will overtake us. The only option here is a global treaty, and knowing what we know of our inability to cooperate to address every other global crisis, such a treaty is a pipe dream. A.I.’s Pandora’s Box has been opened, and there is no way back.

A final objection might be that I am overestimating the capacity of A.I. to keep improving at the dizzying pace of its first breakthrough years. Perhaps … but based on the way other technologies have developed, we don’t really have much reason to believe that. Take computers, smartphones, wearables, cars, or any other major technology we really care about. The range of functions and capabilities just keeps improving year by year, doesn’t it?

What are the likely consequences of A.I.’s displacement of our jobs? Beyond the brute fact of mass unemployment, the first consequence is going to be massive inequality. Income inequality has already been increasing dramatically since the 1970s, but after A.I. kicks most of us out of work, only a few highly compensated owners and superelite workers will be left to rake in huge profit margins from companies that don’t really have to pay salaries anymore. But with the rest of us not working, there will be hardly anyone able to buy what they’re selling. This brings us to the second consequence, which — as many experts have hypothesized — is that giving the unemployed masses some sort of universal basic income is going to be impossible to avoid.

But neither on an individual level nor political level are we ready for such a drastic change. While thinkers from Aristotle to John Maynard Keynes believed additional leisure time could be the key to unlocking our full human potential, we have no reason to believe we are prepared to spend our leisure time productively — or even happily.

Research suggests, in fact, that too much free time — beyond about five hours per day — will tend to decrease people’s life satisfaction, especially if the time is spent in “unproductive activities,” such as watching screens, as opposed to “productive activities,” such as hobbies or exercise. But when given more free time, most people do not use it to exercise or in pursuit of other productive activities and, instead, spend much of it glued to screens.

Other research shows most younger men spend additional leisure time on video games. A study on UBI that was — not-so-coincidentally — funded by OpenAI’s Sam Altman, found that giving 1,000 people an extra $1,000 per month for three years starting in 2020 largely sapped people’s motivation. Those who received the extra cash actually saw income fall by $1,500 per year, as compared to a control group of 2,000 people given a far smaller $50 per month. Moreover, as the researchers reviewing the data concluded:

We find no impact on quality of employment, and our confidence intervals can rule out even small improvements. We observe no significant effects on investments in human capital, though younger participants may pursue more formal education. Overall, our results suggest a moderate labor supply effect that does not appear offset by other productive activities.

None of this means increased leisure time in combination with a well-designed UBI program would not achieve great things in the long run. In a society in which we were taught as children how to spend our time immersed in higher purposes — the endless pursuit of culture and education, spiritual enlightenment, creative endeavors and meaningful social ties against the background of great communities and other occupations of this sort — freedom from the drudgery of work, income to provide for our basic needs and a wealth of leisure time could be a boon to us all.

But we have not been reared in such a society. Worse still, the task before us is made yet more difficult on account of A.I.’s disastrous short-circuiting of the very kind of inner work we will need to perform on ourselves in order to have the skills to spend our leisure time meaningfully. If we are increasingly using A.I. to pre-digest, summarize and interpret great or difficult texts, to write discursively or creatively in our stead and to give us readymade responses to questions that formerly required our ingenuity, engagement, and sustained effort to research, then we are, in the memorable words of Bruno Chaouat in a recent essay in First Things, “outsourc[ing] [our] own inner life to a machine.”

That is a particularly unfortunate state of affairs in a time when we will have to depend on that inner life to be our only redoubt against anomie and ennui.

But we are also not politically ready for this huge paradigm shift. Our staunchly capitalist citizens and leaders are not going to transition seamlessly from mass employment to a society in which most of us are living on the dole. Especially against the background of increasing income inequality and once-thought-untouchable white-collar social elites losing their prime perches at society’s apex, what we have every reason to expect is mass disruption, dislocation, displacement and disaster on every level. Just as we witnessed during the great social cataclysm of the Covid pandemic.

***

The great insight of Edmund Burke, the founder of traditional conservatism — as opposed to the Silicon Valley brand of techno-anarchist libertarian capitalism — is that change must be slow and incremental to preserve the unspoken wisdom embedded in traditional arrangements and institutions. Rapid change, Burke argued, unleashes unforeseen consequences.

Another great philosopher, G.W.F. Hegel, wrote that “the owl of Minerva spreads its wings only with the falling of dusk.” The wisdom, in other words, to comprehend what it is we see unfolding in our midst comes only in hindsight. When we proceed slowly, we can pause to analyze and react in a considered fashion. If ever there was cause, in William F. Buckley’s famous phrase, “to stand[] athwart history yelling stop!,” the advent of A.I. — a technology threatening to displace humanity in every conceivable respect — was it.

Now, it is likely too late for that. But perhaps those of us who have caught sight of the future toward which we are heedlessly careening can at least begin to prepare themselves, their families, and their communities — physically, financially, psychologically, and spiritually — for the earth-shattering consequences.

READ MORE:

Eugenics: The Dark Side of IVF

Artificial Intelligence Requires Human Understanding

The Digital Crucifixion of Christianity