Citrini Research: AI agents could trigger a rapid economic contraction

Citrini Research published a scenario that maps how rapid, economy‑wide adoption of autonomous AI agents could produce a fast-moving macro feedback loop: firms substitute lower-cost in‑house agents for third‑party contractors, which reduces payroll outlays but simultaneously erodes the spending power that sustains other businesses.

In the model’s central run, two headline outcomes frame the risk: a near‑term doubling of unemployment and a greater‑than‑33% drop in aggregate equity value, both concentrated inside roughly a 24‑month window. The mechanism is not simply headcount reduction but the collapse of inter‑firm revenue flows when service spending migrates inside customer balance sheets.

Complementary voices — including industry and policy figures — underscore how that mechanism could be amplified or moderated. Analysts and executives point to outsized capital flows into AI infrastructure (market estimates cited near $1.5 trillion in 2025), which concentrates procurement through a small set of hyperscalers and suppliers and raises vendor‑lock‑in, execution risk and timing exposure for downstream buyers.

Empirical signals already visible in labor and markets strengthen the scenario’s plausibility: tens of thousands of AI‑related layoffs, rising worker anxiety in surveys, and an uptick in investor scrutiny on earnings calls. Market action has, in some cases, preceded fundamentals — with traders de‑risking firms seen as most exposed while sell‑side forecasts lag — producing price moves and wider credit spreads before consistent guidance cuts appear.

Where Citrini centers on B2B substitution (procurement, contract workflows, back‑office services), other expert commentaries widen the frame: some warn that autonomous systems’ speed and scope compress the time available for worker adjustment; others emphasise that concentration of infrastructure spending makes displacement partly an infrastructural problem, not only a managerial one.

The scenario’s timing and magnitude are disputed. Practical constraints — data quality, integration costs, SLAs, legal and liability frictions — mean fully unsupervised replacement is unlikely to be instantaneous; yet semi‑supervised deployments can still scale substitution rapidly enough to produce meaningful demand‑side shocks over quarters rather than decades.

Policy responses highlighted across sources include pairing demand‑side worker supports (retraining, apprenticeship models, targeted cash or stipend pilots) with supply‑side remedies (public investment in open infrastructure, portability and auditability mandates, competition policy to limit hyperscaler lock‑in, and tax or shared‑ownership models to broaden returns).

For corporate leaders and investors the immediate takeaway is operational and diagnostic: map where internal automation will hollow out other firms’ revenues, stress‑test revenue against contractor spending declines, monitor hyperscaler capex and procurement cadence, and incorporate scenario runs where guidance deteriorates within two to four quarters.

In short, the Citrini scenario reframes AI risk from isolated job‑replacement narratives to a systemic macro feedback problem: substitution‑driven demand loss feeding back into profits, investment decisions and asset prices — a pathway made more fragile by concentrated infrastructure spending and by the mismatch between market repricing and formal analyst revisions.