When AI Succeeds, the Economy Suffers: A Disruption We're Not Ready For
I recently came across an article by CitriniResearch that looks back from 2028 at a global economic crisis triggered by AI. While it's a fictional scenario, the logic chain it lays out deserves serious consideration.
Here's our interpretation, critique, and extension of that article.
I. AI's Success Could Be the Economy's Poison
The article describes a counterintuitive scenario: the more powerful AI becomes, the worse the economy gets.
The logic is clear: AI capabilities improve, companies realize AI is cheaper than hiring, so they lay people off. The savings go into more AI, which gets even stronger, leading to more layoffs. Laid-off white-collar workers spend less, corporate revenues suffer, and to protect margins, companies double down on AI and cut more jobs.
A negative feedback spiral with no natural brake.
The article describes how this spiral spreads from tech to the entire economy: SaaS clients downsize and cancel licenses, AI Agents make purchasing decisions for users — killing intermediaries that relied on human inertia — and white-collar unemployment triggers mortgage defaults threatening the $13 trillion mortgage market.
Chilling to read. But upon reflection, this projection has several key flaws.
II. Not That Fast, Not That Simple
Time is compressed. The article squeezes a 5-10 year process into 2 years. In reality, enterprise adoption of new technology is far slower than the technology's own advancement. Compliance reviews, organizational inertia, system integration — these friction forces are real. Humans tend to overestimate short-term impact and underestimate long-term impact.
Intermediaries won't disappear — they'll transform. The article claims AI Agents will eliminate all middlemen. But intermediaries aren't just information relayers — they bear credit and responsibility. When AI can fake everything, "a real human is accountable" itself becomes a scarce commodity. Intermediaries will shift from "information brokers" to "trust brokers."
Human services will become luxury goods. The more AI replaces, the higher the scarcity premium on human services. Therapy, bespoke craftsmanship, live entertainment — prices in these sectors will surge due to scarcity. Total societal wealth hasn't vanished; it's been redistributed.
Business owners aren't algorithms. If every company lays off workers until no one can buy anything, the fear of systemic collapse will trigger collective action at some tipping point — government intervention, industry agreements, or simple self-preservation instinct. The spiral has an equilibrium.
Policy tools are underestimated. During COVID, the US passed trillion-dollar stimulus within weeks. AI taxes, compute taxes, mandatory human employment quotas — these policy tools won't stop the trend, but they'll significantly alter its speed.
New consumption categories are ignored. The internet didn't just kill newspapers — it created social media and streaming. AI may create products and services we can't yet imagine. The article's framework is zero-sum, but technological revolutions are often positive-sum.
III. The Real Problem: Distribution Fracture
Having critiqued the article's extreme framing, the core insight still holds — AI is shaking the foundations of the modern economy.
For two hundred years, the economic cycle has been: people work, earn income, consume, companies profit, hire more people. "Labor" is everyone's ticket to participate in economic distribution.
AI is devaluing that ticket.
GDP is still growing. Output hasn't decreased. But output is no longer distributed to most people through labor. Productivity belongs to capital; purchasing power belongs to labor. When the two decouple, the economic cycle breaks.
This is what we should really worry about — not that AI is too powerful, but that our distribution mechanisms can't keep up.
IV. Deflation, Inflation, or Stagflation?
AI creates deflationary pressure (goods and services become cheaper), but governments must spend massively to support the unemployed (inflationary pressure). What happens when these two forces collide?
The most likely sequence: deflation first, then inflation. AI shock causes economic contraction, then governments are forced into massive money printing. 2008 followed this exact script.
But this time there's a crucial difference.
The traditional money-printing logic: money reaches people, people spend, companies earn revenue, hire workers, cycle recovers. But AI breaks the last step — companies with revenue don't need to hire; they buy more AI. Money flows from government to people to companies, then stops at corporations and compute providers, never flowing back.
Money printing can't create new demand — it can only prevent existing demand from collapsing. Without structural reform, the result is stagflation — prices rise but the economy stagnates. The most painful state.
V. What Can Ordinary People Do?
Become an AI user, not someone replaced by AI. You don't need to learn programming, but you need to embed AI into your workflow. The productivity gap between those who can use Agents and those who can't is already 5-10x.
Stay flexible. Low fixed costs, multiple income streams, continuous learning. The most vulnerable people are those betting everything on "the world won't change" — holding high-paying white-collar jobs, carrying large mortgages, assuming income will grow steadily for 30 years.
Develop skills where AI falls short. Judgment and taste, interpersonal coordination, cross-domain synthesis, narrative and persuasion. AI can generate 100 options, but choosing the right one is human value.
In asset allocation, don't go all-in on any single category. Keep enough cash for optionality during the early stages of a crisis. Long-term, governments will always choose inflation over deflation to resolve existing debt — meaning holding pure cash is a losing strategy. Allocate to assets that hedge against currency devaluation. Compute, energy, and data are the hard currencies of the AI era.
VI. Final Thoughts
Nobody truly knows where AI will take the economy. CitriniResearch's article describes an extreme scenario. Reality will likely be less dramatic, but the direction may be right.
The only certainty: sitting still and waiting for answers is the worst strategy.
Short-term, rely on cash and liquidity to weather the adjustment. Mid-term, hold assets that generate cash flow. Long-term, bet on real demand rather than narratives. Drop the conviction bias, return to first principles, and let the market validate — rather than letting belief predetermine.
The canary is still alive. But it doesn't look comfortable.
This article is based on an extended discussion of CitriniResearch: The 2028 Global Intelligence Crisis. The original article, written from the perspective of June 2028, presents a fictional macro memo reviewing an AI-triggered economic crisis. Highly recommended reading.