A software leveraged buyout in 2024, an annuity sold by a life insurer in 2025, and a thirty-year mortgage written in 2026 look like three separate financial products. They are the same trade. Each one is a bet that white-collar employment in the United States stays roughly where it is. The largest single risk on global balance sheets is not labeled as such anywhere in the disclosures.
This is the daisy chain that makes Citrini Research's 2028 scenario different from a normal cyclical downturn. The displacement story by itself is uncomfortable. The displacement story landing on a financial system whose major credit pools are correlated to the same underlying labor market is something else.
How the bet got distributed
Start with the software side. Hellman & Friedman, Permira, Thoma Bravo, and the rest of the PE software franchise spent the back half of the last decade buying mid-sized SaaS companies on the assumption that recurring revenue is bond-like. Zendesk at $10.2 billion. Citrix. Anaplan. Ping Identity. The leverage works because churn was low and net retention sat above 100% across cycles. That, in turn, is an assumption about white-collar headcount: a customer-service platform's seat count tracks how many agents the buyer employs. When the buyer stops hiring agents because a model handles tier-1 tickets, the bond stops being bond-like.
The annuity side is one step removed and twice as large. Athene, American Equity, Global Atlantic, and a chain of insurer-owned credit franchises spent the same decade selling fixed-rate annuities and reinvesting the float into private credit — direct lending to mid-market companies, including the very SaaS firms in the LBO portfolios. The yield is good because the loans are illiquid and structurally senior. It is also good because regulators rate them on agency models calibrated to past white-collar employment. Apollo, Blackstone, Blue Owl, HPS, and Ares run the same playbook at scale. Life-insurer exposure to private credit is now measured in hundreds of billions of dollars, and the underlying borrowers are the same firms whose customer base is being eaten.
Then the mortgage book. $13 trillion in U.S. residential mortgages, underwritten on stable, high-income employment over a thirty-year horizon. Most of those loans were "good on day one" in a way 2008-vintage subprime never was — real income, real down payments, real credit scores. The exposure is not to underwriting fraud. It is to a future state in which the borrower's job no longer exists.
Why correlation is the killer
Each bet, taken alone, is reasonable. A software LBO assumes its customers keep buying software. An annuity portfolio assumes its borrowers keep paying interest. A mortgage book assumes its borrowers keep their jobs. The hidden problem is that all three assumptions are the same assumption, written three different ways. Stable white-collar employment is the load-bearing input across the entire chain.
In Citrini's scenario, that input deteriorates fast enough that the failure modes line up. ServiceNow's projected ARR growth slows from 23% to 14%, signaling SaaS-wide compression. White-collar job openings fall 15% year over year. By Q3 2027, jobless claims surge to 487K — the highest reading since the early pandemic. Zendesk debt trades at 58 cents on the dollar, not because Zendesk is uniquely broken but because the LBO model that priced it assumed a customer base shrinking slower than principal could amortize. The same dynamic then impairs the private-credit funds that hold the debt, the insurers and pensions that hold the funds, and the mortgage market depending on the same borrowers staying employed at the same kinds of companies.
The 2008 comparison is instructive but inverted. Yale's antimonopoly framework on the AI stack and labor research on the breakdown of entry-level pathways describe a slow grind, not a sudden discovery of fraud. In 2008, the loans were bad on day one and the market took two years to admit it. In 2028, the loans are good on day one and the world changes after.
The question regulators are not yet asking
Regulators stress-test banks against unemployment, rate, and housing-price shocks. None of the major frameworks explicitly stress-test against a structural white-collar displacement that simultaneously impairs SaaS revenue, private-credit cash flows, and mortgage-borrower income. No historical episode contains all three at once. The current technology shock hits all three at once.
Manage risk on any one of these books alone and the position looks fine. The chain is the trade. The honest question for anyone with a balance sheet that touches private credit, SaaS LBOs, or residential mortgages is whether the risk model knows it is holding the same bet three times.