In February 2026, two finance writers named James van Geelen and Alap Shah published a research note on Substack. It wasn't meant to be a prediction. They called it a "thought exercise" — a scenario analysis of what could happen if AI's displacement of white-collar work followed a specific economic logic.
The note went viral. The New York Times covered it. The Wall Street Journal asked who wrote it. And according to multiple reports, it contributed to a sell-off in major tech and financial stocks.
Here's everything that's in it.
The Central Concept: Friction
The report builds its entire argument on one word: Friction.
"The unique ability of human beings to analyse, decide, create, persuade, and coordinate was the thing that could not be replicated at scale. The authors called this historical scarcity Friction."
For centuries, Friction was the moat. You couldn't outsource the doctor's judgment, the lawyer's interpretation, the analyst's pattern recognition. You had to pay a human — and pay them well — because no machine could do it.
The Citrini thesis: AI makes Friction approach zero.
When the thing that made skilled professionals irreplaceable becomes a commodity available to anyone with an API key, the entire economic logic of the knowledge economy breaks.
The Doom Loop
The report isn't just about job losses. It's about a self-reinforcing feedback loop — each step making the next step worse:
↓
② Companies realise they need fewer white-collar workers
↓
③ White-collar layoffs increase — high earners lose income
↓
④ Displaced workers spend less — consumer demand falls
↓
⑤ Margin pressure forces companies to invest even more in AI
↓
⑥ AI capabilities improve further → back to ①
The loop compounds. Each cycle accelerates the next. And unlike previous recessions, the people hit hardest aren't the lowest earners — they're the professionals who never expected to be vulnerable.
Ghost GDP: The Most Important Idea in the Report
This is the concept that made economists sit up.
"Outputs that are shown in national accounts increase — but the money never circulates through the real economy. The authors called this Ghost GDP."
Standard economic models measure GDP as output. AI produces enormous output. So headline GDP numbers look strong. The economy looks fine on paper.
But here's the problem: that output is being generated by AI agents — systems that "don't sleep, take sick days, or require health insurance." The value flows entirely to the owners of the AI infrastructure. It never pays salaries. It never enters households. It never gets spent at restaurants or on mortgages.
Ghost GDP in plain English: Imagine a factory that produces $10 billion of goods using robots owned by five people. GDP says the economy grew. But 10,000 workers who used to run that factory are now unemployed and broke. The five owners are richer. Everyone else is poorer. GDP doesn't capture this distinction — and that's exactly the trap.
The Cascade: How It Spreads Sector by Sector
The report traces the specific industries that fall, in order:
📊 Next wave: Financial advice, tax preparation, routine legal work
👔 White-collar recession: Analysts, consultants, junior lawyers, accountants
📉 Market impact: S&P 500 falls 38% peak-to-trough
📈 Unemployment: National rate hits 10.2%
🏠 Ripple effect: Residential mortgage market destabilised
💸 Financial system: Software company losses → loan defaults → private credit collapse
The key inversion the report highlights: unlike previous technological revolutions where low-skill workers suffered most, this one hits the high earners hardest. High-earning professionals — the ones whose spending props up entire urban economies — get displaced and pushed into the gig economy. Labour supply explodes. Wages collapse across the board.
The "Any Service" Test
"Any service whose value proposition is 'I will navigate complexity that you find tedious' is getting disrupted."
This is the most useful diagnostic line in the entire report. Run any job through this test:
Travel agent: "I'll find the best flights and hotels so you don't have to." → Disrupted.
Tax advisor: "I'll navigate the tax code so you don't have to." → Disrupted.
Financial analyst: "I'll parse the filings and earnings calls so you don't have to." → Disrupted.
Radiologist: "I'll read the scan so you don't have to." → Disrupted.
Any job where the core value is navigating complexity on behalf of someone else is exposed. The question isn't whether AI will do it. It's how fast.
What the Critics Said
The report didn't go unchallenged. The Economist called it economically flawed. The Financial Times called it "extreme and improbable." The main criticism:
The Bull Case Against It
New jobs will emerge. They always have. The report assumes today's jobs are permanent. Every past disruption created new categories of work we couldn't imagine beforehand.
The Bear Case For It
Past disruptions took decades. AI is compressing the transition to years. New jobs may emerge — but not fast enough to absorb displaced workers before the demand collapse triggers the doom loop.
Chinese investors publicly shrugged it off. The South China Morning Post reported China's market was largely unmoved. Whether that's rational skepticism or optimism about their own AI buildout is a different question.
This Is a Scenario. Not a Prediction.
Why It Actually Matters
The reason this report rattled markets isn't the numbers. It's that it gave language to something people were already feeling but couldn't articulate.
"Ghost GDP" is a perfect concept for an era where headline metrics look fine while the lived experience of millions of workers is deteriorating. It explains why a booming AI economy can coexist with widespread anxiety about the future.
The "Friction" concept explains why this disruption feels different from mechanising agriculture or automating manufacturing. The assembly line took physical labour. AI takes cognitive labour — the part that was supposed to be safe.
"The thing that could not be replicated at scale" is now replicable at scale. That sentence contains the entire crisis.
Whether the exact scenario plays out or not, the underlying dynamic is already underway. Travel booking is already dead as a career. Tax prep software already handles 70% of returns. Financial advice is already being democratised by robo-advisors. The cascade has started. The report just extended the logic forward.
The question isn't whether to believe the scenario. The question is: what do you do knowing the direction of travel?
Source: Citrini Research — James van Geelen & Alap Shah, Feb 2026 · Referenced in NYT, WSJ, FT, The Economist, Fortune · Deep dive by Research Hub