Main Topic
The report is a thought‑experiment that imagines a “Global Intelligence Crisis” by 2028: AI and agentic automation radically boost productivity and profits but hollow out white‑collar jobs, crush intermediation business models, and trigger a systemic macro shock where GDP looks strong on paper while the human consumer economy, private credit complex, and even prime mortgages are pushed toward crisis.
Key Insights
The memo is written as if from mid‑2028, with unemployment above 10% and the S&P 500 down 38% from its October 2026 highs, tracing how an AI boom morphed from sector disruption into a full‑blown macro crisis. Early 2026–26: agentic coding tools (Claude Code, Codex, etc.) make it cheap and fast to replicate mid‑market SaaS; procurement teams start threatening in‑house builds to win 20–30% discounts, and long‑tail SaaS (Monday, Zapier, Asana, etc.) gets hit first. Incumbent software vendors like ServiceNow see net‑new ACV decelerate and respond with layoffs and “efficiency programs,” using the savings to buy more AI—each firm’s rational response fuels a collective “intelligence displacement spiral” where improved AI leads to more layoffs, weaker consumption, and more pressure to invest in AI. The key difference from prior tech cycles: incumbents do not resist disruption (like Kodak or Blockbuster) but become AI’s most aggressive adopters because boards demand margin protection when their stocks are already down 40–60%. By 2027, agentic consumer commerce explodes: open‑source agents like Qwen’s shopper run 24/7 on devices, price‑matching and optimizing purchases, driving “friction to zero” in everything from travel booking and insurance renewals to basic legal, tax, and financial‑advice tasks. A massive “rent‑extraction” layer built on human limitations (limited attention, inertia, brand bias, tolerance for bad pricing) is dismantled: subscription CLVs shrink, travel aggregators, brokers, and many advisory models erode, and even real‑estate commissions compress as agents replicate realtor knowledge at scale. The authors define a destroyed moat type—“habitual intermediation”: apps like DoorDash and Uber Eats relied on users being hungry, lazy, and defaulting to the home‑screen app, but agents have no app loyalty and route orders through whichever service is cheapest and fastest, fragmenting markets and crushing margins. Once agents control transactions, they also route around 2–3% card interchange, favoring stablecoins over Solana or Ethereum L2s; this hits Mastercard, AmEx, and card‑focused lenders hard, as rewards and fee‑based models get squeezed. Through 2026, markets misread AI disruption as sector‑specific (software, consulting, payments), but the memo argues the U.S. is fundamentally a white‑collar services economy where about half of employment—and roughly three‑quarters of discretionary spending—comes from the very jobs being automated. The authors argue AI differs from past tech shifts because it can perform new roles as soon as they appear; displaced coders and consultants cannot simply “move up the stack” when AI is already better at those tasks. As AI layoffs mount, displaced white‑collar workers downshift into services and gig work, depressing wages for existing blue‑collar and service workers; an ex‑Salesforce PM driving Uber at a quarter of her prior salary illustrates how upper‑decile income shocks cascade down the income distribution. Consumption damage is amplified because the top 10–20% of earners drive more than half of U.S. consumer spending; a small percentage drop in white‑collar jobs translates into a much larger hit to discretionary demand, with a lag as households first burn savings and draw HELOCs before cutting spending. By late 2027, jobless claims surge and JOLTS data show white‑collar postings collapsing; the economy enters recession even as AI infrastructure (NVDA, TSM, hyperscaler capex, Korea/Taiwan exports) keeps booming. The crisis then jumps from equity to credit: private‑credit funds are stuffed with PE‑backed software and information‑services loans underwritten on the assumption that ARR is truly recurring; as AI automation kills those revenues, Moody’s downgrades billions of software loans and landmark deals like Zendesk’s $5B facility blow up. “Permanent capital” in private credit turns out to be life‑insurance and annuity money routed through insurer–reinsurer webs (Athene, Global Atlantic, etc.); as defaults hit and regulators tighten capital rules, these structures face unanticipated stress and ratings downgrades. The authors describe a “daisy chain of correlated bets on white‑collar productivity growth”: software ARR, consulting fees, card interchange, and mortgage underwriting are all implicitly tied to high and rising professional incomes, which are now in question. The final, most dangerous stage is the mortgage question: a $13 trillion U.S. mortgage market underwritten on the assumption that prime borrowers (780 FICO, 20% down, verified incomes) will stay employed for 30 years is suddenly misaligned with a world where AI has permanently impaired their earnings power. Unlike 2008, loans were “good on day one”; the world simply changed afterward—prime households use credit cards, 401(k) withdrawals, and savings to stay current on mortgages for a while, but are one more shock away from distress, while house prices in tech hubs start to fall and early‑stage delinquencies rise.
Actionable Takeaways
Treat AI as a macro variable, not just a sector theme: assess portfolios and business models for implicit exposure to white‑collar incomes, subscription friction, and intermediation rents that AI agents can compress or eliminate. For investors, be cautious with assets whose cash flows depend on “recurring” software ARR, high‑fee intermediation, card interchange, or prime mortgages in white‑collar hubs; stress‑test scenarios where those revenues prove less durable than history suggests. Favor real‑economy exposure that is less sensitive to agentic disintermediation—hard assets, critical infrastructure, physical production, and jurisdictions whose external balances don’t rely on exporting white‑collar services (India is cited as a vulnerability case). For operators, understand that adopting AI purely as a cost‑cutting tool can collectively erode your own demand base; explore ways to share AI‑driven productivity gains with customers and workers (e.g., pricing, profit‑sharing, or new services) to slow the displacement spiral. Monitor private‑credit and insurer balance‑sheet quality; recognize that “permanent capital” can mask household exposure and regulatory cliff risks when correlated assets are hit by the same AI‑driven shock.
Contrarian Views
The piece challenges the standard “tech creates more jobs than it destroys” narrative, arguing that AI is different because it can instantly move into the new roles that prior tech shifts created for humans. It disputes the idea that AI disruption will stay contained to software/IT, instead presenting a chain in which payments, gig‑economy platforms, card networks, private credit, and even prime housing all share the same underlying vulnerability. The report frames de‑intermediation by agents (killing friction and fees) as structurally deflationary for many business models even as AI infrastructure spending itself remains strong, creating “Ghost GDP” that doesn’t feed human consumption. It flips the usual optimism about “AI as productivity miracle” by modeling a world where abundant intelligence is itself the source of a deflationary, demand‑crushing crisis rather than a simple boon to growth.
Final Thoughts
Citrini and Shah’s “Global Intelligence Crisis” is not a forecast but a stress‑test of the left‑tail: what happens if AI’s upside plays out faster than expected, but the institutional and credit systems built around white‑collar incomes fail to adapt. It forces readers to think beyond sector rotation and ask systemic questions—about private credit, life‑insurance funding, mortgage underwriting, and the fragility of a consumer economy when the consumers themselves are being replaced by machines.
Listen to Full Episode