The AI Jobs Memo That Crashed Tech Markets — And the 10-Phase Crisis Nobody Is Pricing Yet
A research memo went viral on Monday. DoorDash, American Express, and Blackstone had each shed more than 7%. Visa, Mastercard, Salesforce, ServiceNow, Apollo, KKR — all down between 3 and 6 percent. The document responsible was not a regulatory filing or a Fed statement. It was a hypothesis.
Published by Citrini Research and styled as a 2028 retrospective, the memo modeled a world in which AI succeeds so completely it severs the link between corporate productivity and human income — a "Ghost GDP" economy featuring strong nominal output alongside collapsing monetary velocity. Analysts called the sell-off overdone. The memo's authors called it exploratory modeling. Neither response addresses the more disturbing exercise: accepting the premises and following the logic wherever it leads.
What the Memo Already Says
The spiral has no natural brake. AI capability rises, white-collar labor is displaced, consumption falls, margins compress, companies invest further in automation. SaaS vendors collapse first as agentic coding tools let enterprises replicate software internally. Then payments and intermediation die as AI agents price-match continuously and route around transaction fees using stablecoins — dissolving the friction that built Visa and Expedia. Then the gig economy floods with displaced professionals, and the human wage premium evaporates across the board.
The financial shock targets a borrower nobody considered risky: the 780-FICO professional in San Francisco, Seattle, Austin — people whose careers, not loan terms, have been automated away. Thirteen trillion dollars in prime mortgages begins to reprice. Private credit markets, built on the fiction of "recurring revenue," crack as PE-backed software companies default. That is where the memo ends.
That is not where the logic ends.
Phase One: A Pricing Problem for Everything Linked to Human Income
The first thing to break after the memo's scenario is not a specific sector — it is a factor. Markets stop asking whether a premium brand is resilient and start asking whether it is a claim on a shrinking, increasingly precautionary consumer class. The entire aspirational consumer stack derates: travel, home improvement, luxury-adjacent services, private education, suburban service ecosystems. These businesses are not obviously AI-exposed. They are secretly levered to upper-income stability, and that stability is precisely what the memo destroys.
Phase Two: Municipal Finance Fractures
White-collar income is geographically concentrated. When it weakens, so do income tax receipts, property transaction taxes, commercial office taxes, pension contributions, and transit fare revenues — unevenly and irreversibly. Tech-heavy metros deteriorate first. States reliant on high-income earners face structural budget compression. The muni market stops behaving like a market and starts behaving like a credit selection regime, with commodity-anchored and manufacturing-heavy regions pulling away from knowledge-economy cities.
Phase Three: Office Real Estate Reaches Terminal Impairment
The memo focuses on housing. The more direct exposure is office. Fewer workers plus AI-assisted productivity equals permanently lower space demand per unit of output. Class B and C office becomes stranded. Downtown ecosystems — restaurants, parking, retail, transit — deteriorate in sequence. Local and regional banks carrying commercial real estate concentrations emerge as the next latent stress nodes. The crisis transitions from "AI is replacing labor" to "the physical infrastructure built for that labor is overbuilt" — and that is when regional banking stress resurfaces.
Phase Four: Banks Become the Amplifier
The banking system was not the origin of this crisis. It becomes its transmission mechanism. Private credit can hold to maturity; the real economy cannot stop weakening. Deposits slow, CRE losses mount, consumer credit deteriorates among prime-but-stressed households. Banks respond as they always do: tighten standards, cut risk, protect capital. The result is not a bank run — it is credit rationing. And credit rationing is lethal in a transition economy, where small businesses need working capital, households need refinancing flexibility, and municipalities need confidence in debt issuance. The downturn deepens through withdrawal, not collapse.
Phase Five: Expectations Collapse Before the Data Does
The labor shock travels further than unemployment figures capture. Workers who keep their jobs begin behaving like workers who've lost them — before any layoff notice arrives. Raises disappear. Replacement anxiety rises. Household formation slows. Elective healthcare, private schooling, and home upgrade spending contract. Startups outside AI cease to form. The economy enters behavioral austerity among the formerly secure: a demand contraction that precedes the data by years and proves nearly invisible to conventional macro indicators until it is entrenched.
Phase Six: Three Political Blocs, No Clean Majority
The policy response does not arrive as a coherent package. It arrives as a war. Three blocs form with irreconcilable priorities: Protectionists demanding deployment slowdowns, human-in-the-loop mandates, and AI licensing. Redistributors pushing compute taxes, sovereign inference levies, and national AI dividends. Accelerationists insisting on minimal friction and targeted retraining. None commands a majority. The result is a volatile sequence of partial measures, court battles, regulatory fragmentation, and state-federal conflict. That uncertainty alone suppresses investment outside AI infrastructure — which deepens the bifurcation it was meant to prevent.
Phase Seven: Capital Markets Split Into Two Economies
The bifurcation becomes structural. Machine GDP — compute, power, semiconductors, networking, cooling, AI security, machine-to-machine rails — compounds regardless of household demand. Human GDP — headcount-based software, friction-dependent intermediaries, discretionary consumer brands, local services, housing turnover — structurally derates. Indices, blending the two, stop describing reality. The result is massive dispersion, factor instability, and repeated quarters where the macro narrative is wrong while individual stock selection is decisive. Index resilience masks broad pain — until breadth finally breaks.
Phase Eight: Valuation Frameworks Stop Working
The accounting shortcuts that built modern investing quietly fail. "ARR is recurring" — only if customers cannot leave. "Margin expansion is bullish" — only if customers remain solvent. "Prime credit is safe" — only if income paths remain stable. A forensic investing era begins: revenue quality supersedes revenue growth, customer concentration by occupation and income class becomes a core screen, and the critical distinction emerges between moats built on workflow embedding versus moats built on human inattention. In an agentic world, software that persists because switching is inconvenient is not the same as software that persists because it is irreplaceable. Agents do not have inertia.
Phase Nine: Pensions, Endowments, and the Illusion of Diversification
Institutions with long-duration liabilities discover that their portfolios were never truly diversified. Equities, private credit, real estate, and alternative assets all carry the same hidden exposure: the durability of human income. When that single factor deteriorates, correlations converge. Endowments cut spending rates. Pension funding gaps widen. Defined contribution savers approaching retirement face forced de-risking into short-duration sovereigns. Private asset marks face mandatory re-evaluation. The feedback into the real economy — through lower foundation grants, reduced institutional investment, political pressure on pension promises — closes another loop in the spiral.
Phase Ten: Geopolitical Realignment
The international dimension completes the picture. Countries whose current accounts depend on exported services and labor arbitrage — the model Citrini explicitly flags as obsolete for Indian IT — face rising unemployment, FX volatility, and political pressure toward protectionism. Meanwhile, nations controlling the bottlenecks of the AI supply chain — energy, chip fabrication, grid capacity, critical materials — gain structural bargaining power. Export controls and technical standards cease to be national security tools and become macroeconomic policy instruments. Geopolitical fragmentation slows adaptation, raises transition costs, and worsens the outcomes it was meant to manage.
The Master Screen
Across all ten phases, a single investment framework crystallizes. It begins with one question that no index is currently designed to answer: Is this business a claim on human income or machine output?
From there: Does the moat derive from friction, trust, regulation, or genuine irreplaceability? Is revenue recurring because customers cannot leave — or because they haven't yet been given an agent that will leave for them? Is growth coming from real demand expansion, or cost takeout that quietly cannibalizes the customer base?
And finally, the deepest screen of all: how much of any given portfolio's apparent diversification is simply human income beta wearing different labels?
Stabilization, within the memo's own logic, arrives only when policy restores the circular flow the displacement destroys — through transfers, compute taxation, adapted debt contracts, or some socialization of AI rents sufficient to reconnect production to demand. None of these are easy. None have a majority. None are priced.
The sell-off on Monday will likely be forgotten by Friday.
The question it briefly forced into the open — what institutions only work if human intelligence remains scarce? — will not be so easy to close.
not investment advice