The AI scare trade started in software. Then it hit insurance. Then wealth management. Then payments. Then IT outsourcing. The market isn't pricing one thesis. It's pricing disruption one product launch at a time.
On February 9, 2026, a company called Insurify launched an app in OpenAI’s ChatGPT directory. It lets users compare auto insurance quotes through natural conversation, drawing on 196 million historical quotes and 70,000 verified reviews.
Willis Towers Watson fell 12% that day — its worst session since November 2008. Arthur J. Gallagher lost 9.9%. Aon dropped 9.3%. The S&P 500 Insurance index had its biggest decline since October.
The next day, Altruist Corp. unveiled Hazel, an AI tax planning tool that gives financial advisors real-time scenario modeling — the impact of a bonus, a home sale, a retirement transition — updated instantly. LPL Financial fell 8.3% after touching 11% intraday. Charles Schwab lost 7.4%. Raymond James dropped 8.8%. Altruist’s CEO said the tool makes “average advice a lot harder to justify.”
Two weeks later, on February 23, Anthropic announced that Claude Code can modernize COBOL, a language from 1959 that still processes trillions of dollars in daily financial transactions on IBM mainframes. IBM fell 13.2% — its worst day since 2000.
Then came the Citrini report.
The Scenario That Moved Markets
On February 22, Citrini Research published “The 2028 Global Intelligence Crisis” on Substack. It was framed as a thought experiment — a hypothetical macro memo written from June 2028, looking back at an AI-driven economic downturn. The authors explicitly called it a scenario, not a prediction.
The market treated it as a prediction anyway.
Software and payments stocks slid on Monday. Delivery platforms followed. Private capital firms. Financial services companies. The Dow dropped 822 points. Indian IT stocks had their worst session since August 2023, with the Nifty IT Index falling 5.3% as Citrini’s scenario projected contract cancellations for firms like Tata Consultancy, Infosys, and Wipro accelerating through 2027.
What made the report land was not its novelty. Every claim in it had already been circulating. AI replacing software engineers. AI compressing consulting engagements. AI eliminating per-seat pricing models. These are February’s conventional wisdoms.
What made the report land was its structure. It connected the sectors into a single narrative. It showed how one displacement feeds into the next: companies replace workers with AI, displaced workers spend less, weakened demand leads to more AI investment to protect margins, and the cycle tightens. Each sector’s disruption accelerates every other sector’s disruption.
The market had been pricing AI disruption one stock at a time. The report showed the cascade.
The Product-Shaped Repricing
The pattern that emerged in February is structurally different from previous technology selloffs.
The dot-com crash was thesis-driven. “The internet changes everything” became “these companies have no revenue,” and valuations collapsed together. The 2022 rate-driven selloff was mechanical — rising rates compressed all growth multiples simultaneously. Both were broad. Both were abstract.
The AI scare trade is neither. It is product-shaped. Each selloff has a specific trigger: a named AI tool that does a named thing that a named company currently charges for.
Insurify’s app compares insurance quotes. Willis Towers Watson helps clients compare insurance quotes. The app launched. The stock fell. The connection is not theoretical. It is a straight line from product to price.
Altruist’s Hazel does real-time tax scenario modeling. LPL Financial’s advisors do real-time tax scenario modeling. The tool launched. The stock fell.
Claude Code modernizes COBOL. IBM charges billions to modernize COBOL. The blog post published. The stock fell.
This is market discovery operating at the product level, not the thesis level. The question is no longer “will AI disrupt this sector?” It is “which specific product will do it, and has it shipped yet?” Each sector waits for its own Claude Code moment — the day a specific tool makes a specific revenue line compressible.
Ghost GDP
The Citrini report introduced a concept worth naming separately: ghost GDP.
The scenario projects labor’s share of GDP falling from 56% to 46%. Output grows — AI makes everything more productive. But the money shows up in national accounts without circulating through the real economy. Productivity gains accrue to capital owners. Displaced households don’t spend. GDP rises while consumption falls.
This is the structural version of the AI scare. It is not that AI destroys value. It is that AI concentrates value — the economy produces more while fewer people participate in the production. The aggregate numbers look fine. The distribution underneath does not.
The ghost GDP concept explains why the scare trade hits consumer-facing sectors hardest. Payments companies need consumers spending. Insurance brokers need consumers buying policies. Wealth managers need consumers accumulating wealth. If the income that funds consumption shifts from labor to capital, every business model built on broad consumer participation weakens — even if the economy as a whole grows.
Alap Shah, co-author of the report and CIO of Lotus Technology Management, went on Bloomberg TV to call for an AI tax — a levy on windfall gains from AI productivity to cushion the employment transition. The proposal immediately became the most discussed policy response since the selloff began.
The Contagion Map
Step back and look at the timeline:
Early February: the SaaSpocalypse. More than $800 billion erased from enterprise software. Abstract fear — AI might replace developers, could collapse per-seat pricing. Broad, diffuse, thesis-driven.
February 9: insurance brokers. One product, one stock, one day. Insurify launches. WTW falls 12%.
February 10: wealth management. One product, one stock, one day. Altruist launches Hazel. LPL falls 8%.
February 23: IBM. One blog post, one stock, one day. Claude Code does COBOL. IBM falls 13%.
February 23–24: the Citrini report connects everything. The market stops pricing disruption piecemeal and starts pricing the cascade. Dow drops 822 points. Indian IT collapses. Payments, delivery, private credit all sell off together.
The contagion is not random. It follows the structure of the economy. Each sector that gets hit reveals the next sector’s exposure. Insurance brokers fall, and the market notices that wealth managers face the same disintermediation. Wealth managers fall, and the market notices that any business built on complexity premiums — where the difficulty of the work is the product — is structurally exposed.
The cascade is the market mapping which revenue lines are complexity premiums, and which AI tools are compressing those premiums, in real time.
What the Cascade Reveals
The most striking feature of the AI scare trade is what it is not.
It is not irrational. Each selloff has a specific, verifiable trigger. The insurance comparison app works. The tax planning tool works. The COBOL modernizer works. The market is not panicking about hypotheticals. It is repricing named capabilities.
It is not uniform. Asian stocks are outperforming, because Asian economies are more manufacturing-heavy and less services-heavy. The cascade follows the structure of the economy it’s hitting. Countries with large white-collar service sectors — the US, UK, India — are most exposed. Countries that make things rather than advise on things are less exposed. The geography of the selloff maps the geography of the complexity premium.
And it is not over. The Citrini report lists sectors that haven’t been hit yet: real estate, travel booking, consulting. Each waits for its product-shaped trigger — the AI tool that does the specific thing they currently charge for. The question is not whether those triggers will arrive. It is when.
The cascade does not require AI to succeed everywhere. It only requires AI to succeed in enough specific places that the connected sectors lose confidence simultaneously. A Substack thought experiment about 2028 moved markets in 2026 not because it was right, but because the market had already seen enough individual data points to believe the pattern was real.
Each product launch is a coordinate. Each selloff is a connection. The cascade is the market drawing the map.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)