Why AI Will Lead to MORE Hiring - and Why Everyone (Yes, Everyone) Is Wrong About It
Citrini's viral AI doomsday memo nuked software stocks in February. It was clever, it was well-written, and it was dead wrong. Profit margins drive hiring. They always have. AI is about to potentially hand corporates the biggest margin windfall in modern history, and the response will be the same as it has been for a century: more jobs, not fewer. The bloodbath isn't coming. The hiring boom is.
Back in February, Citrini Research published a 7,000-word piece of financial fan fiction called “The 2028 Global Intelligence Crisis”, a “memo from the future” in which AI agents vaporize the white-collar workforce, unemployment prints 10.2%, and the S&P is down 38%. It said “scenario, not a prediction” in the second sentence, which naturally meant everyone traded it as a prediction. Software stocks got clubbed like baby seals, IBM had its worst day since 2000, and a top White House economist had to go on record calling the whole thing “science fiction.”
Mission accomplished for Citrini: maximum fuzz, minimum accountability. The problem? The thesis is 100% wrong, and not in a subtle way. It's wrong in the way that anyone who has ever sat in a corporate budget meeting knows it's wrong. Here is my essay-style attempt at explaining why, with charts, anecdotes, and a healthy dose of disrespect for the doomsday-industrial complex.
Warning: The discussion below assumes that AI ultimately makes everything cheaper. That is not true as of today, but it is a reasonable assumption for the next cycle.
Part 1: Hiring follows profit margins. It always has. It always will.
Let's start with the most boring, most reliable relationship in all of macro: when profit margins go up, hiring goes up. Not sometimes. Not “in the pre-AI paradigm.” Always. Four decades of NABE survey data, one conclusion: there is no such thing as a profit boom without a subsequent hiring boom.
Chart 1: Profit margins lead employment like night follows day. Four decades of evidence.
Why? Because organizations are not spreadsheets. They are living organisms with the self-restraint of a labrador at a pedigree pal buffet. When budgets are met and exceeded, jobs materialize out of thin air. Every middle manager on planet Earth knows that headcount equals status. Nobody ever got promoted for shrinking their team; you get promoted for “building out the function.” Call it the kingdom fallacy: give a manager a margin windfall and he will not return it to shareholders. He will hire three strategy leads, a chief of staff, and someone whose entire job is to make slides about the other four.
So walk through the Citrini logic with me: AI works brilliantly, margins explode higher... and then companies fire everyone? That has literally never happened in the history of corporate capitalism. Fat margins burn holes in managerial pockets. If AI delivers the margin expansion the bulls promise, the empirical, boring, four-decade-old response is: hiring goes UP.
The Jevons Paradox, or: why cheaper intelligence means MORE intelligence
In 1865, William Stanley Jevons noticed something odd: as steam engines got more coal-efficient, Britain burned more coal, not less. Efficiency made coal-power cheap, cheap made it ubiquitous, ubiquitous meant total consumption exploded. Now swap “coal” for “white-collar output.”
Chart 2: A sketch, not a prediction. Unlike some research shops, we label our fiction.
When the cost of producing a legal memo, a marketing campaign, or a piece of code collapses by 90%, demand for legal memos, campaigns and code does not stay constant. It explodes. Companies that could afford one analysis will want fifty. Firms that never had a legal department will get one. The bottleneck moves to the humans who direct, verify, sell, and clean up after the machines.
And this is not theory. The flood has already started. Take software: GitHub now sees more than 230 new code repositories created every minute, added 36 million new developers in a single year (that's one per second), and logged close to a billion commits in 2025, up 25%. The marginal cost of building an app has collapsed, so everyone and their barista is shipping apps now:
Chart 3: Coal, but for code. Cheaper production, exponentially more of it.
Books, same story, only funnier. Self-publishing was already booming, and then ChatGPT arrived and things got so out of hand that Amazon had to cap authors at three books per day, a totally human writing pace. Roughly 1.4 million self-published titles now hit Kindle every year, and the trajectory is straight up:
Chart 4: Jevons in hardcover. When writing a book takes an afternoon, everybody writes a book.
Notice what these examples tell you… AI doesn't shrink output, it multiplies it until we are drowning in supply. Apps, books, songs, marketing copy, legal boilerplate: infinite. And when supply is infinite, the scarce (and employable) skill becomes sorting, curating, verifying, and selling it. That, dear reader, is a human job, and we're going to need a lot of them.
Meanwhile, the AI executives keep promising apocalypse with the confidence of men whose valuations depend on it. Anthropic's Dario Amodei warned AI could wipe out half of all entry-level white-collar jobs and push unemployment to 10-20% within five years. Sam Altman muses that whole job categories will “totally go away.” Microsoft's AI chief calls the technology “fundamentally labor replacing.” Gentlemen, respectfully: we've heard this one before. ATMs were supposed to kill bank tellers, yet teller employment roughly doubled between 1970 and 2010, because cheap branches meant more branches. The spreadsheet was supposed to kill accountants. Instead it killed the bookkeeper's pencil and spawned millions of analyst jobs. In the 19th century, 98% of the labor required to weave cloth was automated, and the number of weaving jobs went up, because cloth got cheap and everyone suddenly wanted four shirts instead of one.
The AI-executives are not making an economic argument. They're rehearsing the investor presentation.
The Novo Nordisk case study: hiring follows profits, even when it shouldn't
If you want a live-action demonstration of the kingdom fallacy, look no further than my home country's crown jewel. When the “fat drug” turned Novo Nordisk into Europe's most valuable company (at one point worth more than Denmark's entire GDP), then-CEO Lars Fruergaard Jørgensen went on the hiring spree of the century. Headcount ballooned from roughly 48,000 to over 78,000 in about four years. Did Novo need 30,000 extra people to manufacture a molecule discovered years earlier? Of course not. But the margins were there, so the org chart grew: project managers, sustainability coordinators, internal communications partners, people whose job it was to coordinate the coordinators.
We know it was empire-building rather than necessity, because the moment competition from Eli Lilly bit into margins, new CEO Mike Doustdar cut 9,000 jobs (the largest layoff in Danish history), citing the need to “reduce complexity” after “a period of hyper-growth in employee numbers.” Translation: we hired thousands of office workers because the money was burning a hole in our pocket. Meanwhile Eli Lilly generated roughly double Novo's GLP-1 sales with a leaner setup and never had to stage a Danish national trauma. Profits up, hiring up; profits squeezed, hiring down. The causality runs through the P&L, not through the technology.
And Novo is not special. Google's headcount tripled in the 2010s while its core product needed, generously, a few thousand engineers to maintain. Twitter famously lost 80% of its staff and the app... kept working. The lesson isn't that those workers did nothing. It's that profitable companies hire far beyond any measurable “need,” because that is what profitable organizations do.
Part 2: The corporate pseudo-job: nature's most resilient species
Which brings us to the second thing Citrini's model can't handle: organizations are pseudo-job factories. The modern corporation contains entire ecosystems of roles whose connection to output is, let's say, spiritual. The internal workshop facilitator. The transformation office. The DEI department that grew 55% in a single hiring boom and then had a “strategic realignment” two years later. The eleven-person team preparing the offsite about improving cross-functional synergies, which produces a deck, which is presented at another offsite.
If AI were going to eliminate jobs that don't produce measurable output, it would have to explain why those jobs exist now, in a world of ruthless quarterly capitalism. The answer is that firms don't hire to the productivity frontier. They hire to the budget frontier. Exhibit A, the American healthcare system:
Chart 5: Since 1970, US physicians +~150%. Healthcare administrators +~3,800%. The scalpel has never been outnumbered like this.
Administrative staff now make up roughly a quarter of the entire US healthcare workforce, and administration eats close to a third of every healthcare dollar, about double the share of any other rich country. None of these people existed in 1970. Technology, regulation and revenue created them out of thin air. The pattern repeats everywhere: universities added deans and deputy-vice-provosts far faster than professors; banks added compliance officers by the stadium-full after 2008 (JPMorgan alone added over 13,000 in a couple of years).
Every productivity revolution in modern history has been absorbed by an army of new coordinators, checkers, and workshop-holders. AI will produce compliance-of-AI officers, prompt governance councils, and Heads of Agent Experience. I am only half joking, and the half that's joking is nervous.
And lest you think this is a purely American disease, allow me to bring it home to Denmark. The number of midwives employed by the Danish regions rose 57% between 2007 and 2020. The number of births? Down roughly 5%. More midwives, fewer babies, and somehow still a permanent national debate about midwife shortages. It is the purest data point I know: healthcare hiring detached itself from healthcare output decades ago, on both sides of the Atlantic:
Chart 6: 57% more midwives to deliver 5% fewer babies. Somebody is inventing jobs, and it isn't the babies.
This is, frankly, a European superpower. Nobody on Earth is better at inventing our own jobs than we are, and once the job is invented, at spending the workday on everything except the thing we were nominally hired to do. Committees about the work. Seminars about the committees. An away-day to align on the seminar. The org chart grows, the output doesn't, and everyone agrees the real problem is understaffing. When AI arrives to do the actual task, Europe's white-collar class will be just fine: we stopped doing the actual task years ago.
Part 3: The great reshuffle: buy flexibility, sell rigidity
None of this means nothing changes. Jobs will be reshuffled on an epic scale. The composition changes even as the total grinds higher. And here's the actionable macro bit of this essay -> the winners will be economies where reshuffling is legal.
Chart 7: Employment protection across major economies. AI gets “swallowed” fastest where hiring and firing is easiest.
AI adoption requires churn: kill role A on Monday, create role B on Wednesday. In the US, UK, Canada and Scandinavia's flexicurity systems (Denmark: easy to fire, generous to the fired, the labor-market equivalent of a firm handshake and a warm blanket), that's a Tuesday. In the Netherlands or Italy, firing someone is a multi-year legal odyssey involving courts, works councils, and possibly the Pope. Rigid labor markets won't be spared the disruption. They'll just get it slower, with higher structural unemployment and none of the upside. Add a big, agile service sector and the ranking writes itself: the US and the Anglosphere swallow AI whole; continental Europe will most likely chew on it for a decade.
One honest caveat to my own bullishness: the bar for employment gets raised. When the productivity hurdle rises, some workers don't clear it. Notice that US employment-to-population peaked around the year 2000: introduce internet, peak jobs:
Chart 8: US employment-to-population peaked with the internet. The rest of the West has actually held up better.
Interestingly this is a fairly American disease. European employment rates have kept setting record highs, partly because flexicurity systems retrain people instead of parking them. But the direction of travel is clear: more hiring overall, yet a fatter tail of people permanently on benefits because the minimum productivity bar moved above them. Whether we call the fix “UBI” or something more politically digestible, the transfer state grows. That's a fiscal story (bearish bonds, if you ask me), not an employment apocalypse.
Part 4: The dot-com comparisons: seductive, lazy, and mostly wrong
Every strategist above 50 has one crutch: “it's just like 1999.” It isn't, and the differences all point the same direction. Start with adoption speed:
Chart 9: Generative AI reached 40% of US adults in ~2 years. The internet needed the better part of a decade.
Generative AI is diffusing 4-5 times faster than the internet did. In 1999, we were pricing a revolution that hadn't reached most households. Pure hope, zero cash flow. Today the revolution reached your aunt before the analysts finished their initiation reports. That's why this cycle is earnings-driven, not eyeballs-driven: the hyperscalers funding the buildout are the most profitable companies in the history of money, paying for capex out of operating cash flow rather than vendor financing and dreams.
And the leverage picture? In 2000, margin debt as a share of market cap was spiking into the peak; speculative borrowing was the rocket fuel. Today it sits well below the old highs. The fuel this time is retained earnings, which burn slower and don't explode when the Fed sneezes:
Chart 10: Margin debt to S&P 500 market cap. Notice what's missing: the 2000-style leverage spike.
Which brings me to my favorite genre of financial commentary: the Nominal World People. “RECORD CAPEX!” they shout, pointing at a dollar figure that is a record because every dollar figure is a record in a growing nominal economy. Their grocery bill is also a record. Scale it properly and the hysteria deflates:
Chart 11: The railway barons spent 6-7% of GDP. The AI buildout is ~2%. Some bubble.
Britain's railway mania consumed ~7% of GDP. American railroads ate ~6%. The AI datacenter buildout, for all the breathless headlines, is running around 2% of GDP. Bigger than telecom in 2000, sure, but financed by profits instead of junk bonds. If this is a bubble, it's the most solvently funded bubble in industrial history.
Conclusion: Everyone is wrong, and that's bullish
Conveniently the data has already started delivering on the point I am trying to make in this essay. Apollo's firm-level work shows that companies adopting AI most intensively aren't cutting entry-level staff. They're hiring more of them, dramatically so, within two years of adoption..
Chart 12: High-intensity AI adopters INCREASE entry-level headcount. The doom trade's inconvenient chart.
So let's assemble the machine: AI lifts margins → margins have always, without exception, driven hiring → managers empire-build because that is their biological imperative → Jevons multiplies demand for the newly cheap output → pseudo-jobs bloom in the fiscal sunshine → flexible economies reshuffle fastest and win biggest. That is a positive, self-reinforcing loop: profits fund jobs, jobs fund consumption, consumption funds profits. It is almost embarrassingly conventional, which is exactly why it will be right.
Citrini's memo was designed to create a fuzz, and credit where due: it worked beautifully. Software stocks tanked, Citadel wrote a rebuttal, The Economist weighed in, and a thousand fund managers got to feel intellectual for a week. But being viral and being right are different sports. The doomsday scenario requires corporations to behave in a way corporations have never behaved: handed the potential biggest margin windfall in modern history, and responding with monastic restraint. I'll take the other side of that trade all day.
More profits. More hiring. More pseudo-jobs than any of us can imagine. See you at the internal workshop on agent governance. Attendance is mandatory, and yes, there will be a pre-read deck…
Andreas Steno, CIO of ASMR Wealth, Founder of Nowcast IQ and Macro Strategist at Real Vision
