R
RESEARCH HUB
Intelligence Brief
Macro AI Scenario Analysis

The 2028 Global Intelligence Crisis

CitriniResearch's viral macro scenario that moved markets and sparked a global debate: what happens when AI is wildly successful — and that success destroys the economy that funded it?

What you're reading

This is a scenario, not a prediction. CitriniResearch wrote it as a thought exercise — a macro memo from June 2028, looking backwards at how the economy collapsed. The conceit: what if AI is everything the bulls say it will be, and that's still bad? It went viral, moved markets, and ended up in the New York Times, Wall Street Journal, and Financial Times. Here's the full breakdown.
The Setup

The Premise: What If AI Wins?

The report opens in June 2028 with a 10.2% unemployment print — a 0.3% upside surprise. Markets sold off 2%, pushing the S&P's cumulative drawdown to 38% from its October 2026 highs. Traders, the memo notes, have grown numb.

"Two years. That's all it took to get from 'contained' and 'sector-specific' to an economy that no longer resembles the one any of us grew up in."

The scenario doesn't begin with catastrophe. It begins with euphoria.

By October 2026, the S&P flirted with 8000. Nasdaq broke 30k. Productivity was booming. Real output per hour rose at rates not seen since the 1950s. Corporate profits were record-setting. AI was working — really working.

The only problem: the economy wasn't.

Ghost GDP

Ghost GDP: The Core Concept

This is the intellectual heart of the piece. And it's simple once you see it.

Ghost GDP — output that shows up in national accounts but never circulates through the real economy. A single GPU cluster in North Dakota generating the work of 10,000 white-collar workers in Manhattan is, as the authors put it, "more economic pandemic than economic panacea."

The economy kept growing on paper. But the question nobody asked was: how much money do machines spend? The answer is zero. Machines don't buy cars. They don't pay rent. They don't go out to dinner. AI agents that don't sleep or require health insurance produce output — but that output doesn't become income for anyone who then spends it.

The velocity of money flatlined. The human-centric consumer economy, 70% of GDP, began to wither.

The Spiral

The Intelligence Displacement Spiral

The report identifies a feedback loop with no natural brake:

AI improves

Companies need fewer workers

White-collar layoffs increase

Displaced workers spend less

Margin pressure pushes firms into more AI

AI improves further

repeat indefinitely

White-collar workers saw their earnings structurally impaired. Their incomes were the bedrock of the $13 trillion mortgage market. Prime mortgages, previously considered safe, needed to be reassessed. Private credit markets, bloated with PE-backed software deals that assumed recurring revenue would stay recurring, began to crack.

How It Started

How It Actually Started: SaaS First

In late 2025, agentic coding tools took a step-function jump. A competent developer with Claude Code or Codex could replicate a mid-market SaaS product in weeks. Not perfectly — but well enough that CIOs reviewing $500k annual renewals started asking: what if we just built this ourselves?

The pivot moment: a procurement manager at a Fortune 500 told a SaaS salesperson he'd been in conversations with OpenAI about replacing the vendor entirely with internal AI builds. They renewed at a 30% discount. That, he noted, was a good outcome. The long tail of SaaS, Monday.com, Zapier, Asana, had it worse.

📉 ServiceNow Q3 2026: Net new ACV growth decelerates 23% → 14%
📉 Announces 15% workforce reduction + "Structural Efficiency Program"
📉 Shares fall 18% | Bloomberg, October 2026

The company selling workflow automation was being disrupted by better workflow automation — and its response was to adopt the technology disrupting it.

This was the historical anomaly. The Kodak model — incumbent resists, loses to nimble entrants, dies slowly — didn't apply. These incumbents couldn't afford to resist. With stocks down 40-60% and boards demanding answers, AI-threatened companies did the only rational thing: cut headcount, redeploy savings into AI, use AI to maintain output at lower cost.

"Each company's individual response was rational. The collective result was catastrophic."
When Friction Died

When Friction Went to Zero

By early 2027, LLM usage became default — the way autocomplete or spell-check are default. People used AI agents without knowing what an agent was.

The catalyst for the next phase: Qwen's open-source agentic shopper. Within weeks, every major AI assistant had integrated agentic commerce. Distilled models ran on phones and laptops. The marginal cost of inference collapsed. Consumer decisions stopped being discrete human choices and became a continuous optimization process running 24/7.

By March 2027, the median American was consuming 400,000 tokens per day — 10x since end of 2026.

The concept the report introduces here is Friction. The entire post-industrial economy was built on it:

The U.S. economy built a giant rent-extraction layer on top of human limitations. Things take time. Patience runs out. Brand familiarity substitutes for diligence. Most people accept a bad price to avoid more clicks. Trillions of dollars of enterprise value depended on those constraints persisting. When AI removed them, all of them, those trillions evaporated.

Casualties of Zero Friction

Travel platforms — gone first, because they were the simplest. Agents assembled complete itineraries (flights, hotels, ground transport, loyalty optimization, refunds) faster and cheaper than any platform.

Insurance — the entire renewal model depended on policyholder inertia. Agents eliminated inertia.

Financial advice, tax prep, legal work — any service whose value proposition was "I will navigate complexity that you find tedious" was disrupted.

Subscriptions — passively renewing despite months of disuse. Agents renegotiated or cancelled them.

Retail — humans don't have time to price-match across five platforms. Machines do. Customer lifetime value — the metric the entire subscription economy was built on — structurally declined.

The Cascade

The Financial Cascade

Software defaults began mid-2027. But the disruption didn't stay contained to software. By end of 2027, it threatened every business model predicated on intermediation.

The system turned out to be one long daisy chain of correlated bets on white-collar productivity growth. When that growth inverted, the chain unwound all at once. The November 2027 crash accelerated every negative loop already in motion.

📊 S&P 500 peak: ~8,000 (October 2026)
📊 S&P 500 trough: ~4,960 (38% drawdown by June 2028)
📊 Unemployment: 10.2% (June 2028)
📊 $13 trillion mortgage market: under reassessment
📊 PE-backed software: wave of defaults
📊 Private credit: systemic risk emerging
The Analogy

The Core Analogy: Horses

The report uses one analogy to explain everything. It's the most important paragraph in the piece.

"Horses were not put out to pasture when the automobile arrived. They were put to work differently — until they weren't. And then the economy that had been built around them, feed, stables, farriers, trainers, veterinarians, all contracted. The question isn't whether humans are horses. The question is whether the economy built around human cognitive labour will contract the same way."

The horse economy didn't collapse because horses became worthless. It collapsed because the system built around horses — every job, every business, every asset price that assumed horses would be the dominant form of transport — suddenly didn't need to exist.

White-collar cognitive work is the horse. The question isn't whether humans can still think. The question is whether the economy needs them to — and whether it was built on the assumption that they would.

The Counterargument

Why Critics Say It's Wrong

The Economist and the Financial Times both published responses calling the economics "extreme and improbable." Their main arguments:

1. New jobs always emerge. Every technological wave destroyed jobs and created new ones. Agricultural to industrial. Industrial to service. There's no reason to think this transition is different.

2. Lower prices increase real wealth. If AI makes goods and services dramatically cheaper, the real purchasing power of even lower wages increases. Deflation isn't inherently bad.

3. The timeline is compressed. The scenario assumes everything happens in two years. Real economic transitions take decades. Regulatory friction, adoption curves, and institutional inertia slow everything down.

The authors' response: They called it a scenario, not a prediction. Their goal wasn't to say "this will happen." It was to model a left tail risk that had been almost entirely unexplored. The point is preparation, not prophecy.

What It Means

What This Actually Means for Education

The scenario has a specific, urgent implication for anyone thinking about the future of education and learning.

The jobs that drove demand for higher education — the jobs that made a degree worth $200k in student debt — were predominantly white-collar cognitive work. Analysis. Decision-making. Document production. Financial reasoning. Legal interpretation. These are precisely the jobs the scenario describes being disrupted first and fastest.

If the economic returns on these skills collapse, the entire model of education-as-investment collapses with it. Not because learning stops having value. But because the credential-to-income pipeline that made university the obvious choice breaks down.

The people who will navigate this, whether the scenario plays out fully or partially, are the ones who treated their education as genuine capability development rather than credential acquisition. And those who build or control the intelligence infrastructure will benefit regardless of what happens to everyone else.

This is the thing CitriniResearch is actually saying: the owners of compute will be fine. Everyone else is a variable in their productivity calculation.


Source: CitriniResearch — The 2028 Global Intelligence Crisis · James van Geelen & Alap Shah · Feb 22, 2026 · 171K subscribers · Covered by NYT, WSJ, FT, The Economist, Fortune · Deep dive by Research Hub

Back to Intelligence Brief

Continue Reading

The Fuckening. 6 min → Stability Is the New Status. 6 min →