Overview
In February 2026, CitriniResearch released a special report co-authored with Alap Shah (founder of LOTUS): "The 2028 Global Intelligence Crisis" — a fictional macro memo dated June 30, 2028. The authors explicitly state this is not a prediction but a thought exercise, exploring a relatively neglected scenario: what if continued AI bullishness turns out to be correct, yet the result is bearish? The core question: AI succeeds, but the economy breaks.
- Original article: https://www.citriniresearch.com/p/2028gic
- Published: February 22, 2026
- Intelligence Displacement Spiral: AI capability improves → firms lay off white-collar workers → the unemployed cut consumption → firms invest more in AI to protect margins → AI improves further. An accelerating loop with no natural brake.
- Unlike past tech revolutions: ATMs and the internet destroyed some jobs but created new ones for humans. AI replaces *intelligence itself* — and can already manage AI — so displaced workers cannot simply move into "AI management."
- Labor share of GDP (fictional): 64% (1974) → 56% (2024) → 46% (2028) — the steepest decline on record.
- Travel booking: AI compares all platforms simultaneously; loyalty goes to zero.
- Insurance: annual AI re-shopping wipes out 15–20% of premium income.
- Real estate: 5–6% commissions built on information asymmetry compress below 1%.
- Delivery (DoorDash case): competitors clone the app in weeks, drivers run multi-platform dashboards, agents always pick the cheapest option — margins compress toward zero.
- Finance: "complexity handling" fails because AI doesn't find things complex.
- Revenue falls (payroll and income taxes) while spending rises (unemployment, retraining); automatic stabilizers assume unemployment is temporary, but this displacement may be permanent.
- Rate cuts fail: the problem isn't tight financial conditions but AI devaluing human intelligence.
- QE fails: buying MBS doesn't change the fact that a $200/month Claude agent replaces a $180K product manager.
- Fiscal stimulus requires more transfers from a shrinking tax base, amid political deadlock: the right opposes transfers and warns that taxing compute cedes ground to China; the left fears regulatory capture; hawks cite unsustainable deficits while doves cite post-2008 premature tightening.
- For investors: audit exposure to assets premised on scarce human intelligence — SaaS, intermediated services, private credit; consider hedging between "AI infrastructure" and "AI victims."
- For policymakers: watch labor-market *structure*, not just headline unemployment; rethink the social contract (e.g., proposed taxes on AI output).
- For individuals: build skills hard for AI to substitute (creativity, empathy, complex judgment), stay adaptable, and maintain a financial buffer.
Key points
The fictional 2026–2028 collapse timeline
Early 2026: Software breaks
Agentic coding tools (e.g., Claude Code, Codex) let one skilled developer replicate a mid-sized SaaS product's core features in weeks. Firms ask why pay $500K for software when they can build it with AI. Long-tail SaaS (Monday.com, Zapier, Asana) is hit hardest. In the memo, ServiceNow's Q3 2026 net-new ACV growth falls from 23% to 14%, it announces 15% layoffs, and the stock drops 18%.Mid-2026: The forced self-replacement paradox
"Companies threatened by AI become AI's most aggressive adopters." Unlike Kodak or Blockbuster's slow decline, this is *accelerated self-substitution*.2027: The intermediation layer dissolves
AI agents handle consumer decisions, destroying businesses built on human friction:Mid-2027: Financial system stress
Private credit loans were underwritten assuming ARR is recurring; AI makes ARR merely "revenue that hasn't left yet." Zendesk becomes the largest default: $5B in loans marked to 58 cents. Meanwhile insurers owned by Apollo, KKR, and Blackstone as funding vehicles face capital-ratio forced selling as underlying loans default — "permanent capital" turns out to be ordinary households' annuity savings.2028: Mortgage crisis
$13 trillion in mortgages rests on the assumption that borrowers stay employed. San Francisco home prices fall 11%, Seattle 9%, Austin 8%; early defaults appear in high-income ZIP codes (>40% tech/finance employment), including 780+ FICO "prime" borrowers.> Key insight: 2008's loans were bad on day one; 2028's loans were good on day one — the world just... changed.
The policy dilemma
Warning — but not destiny
> "For the first time in history, the most productive asset in the economy produced fewer jobs rather than more."
Our frameworks were designed for a world where scarce inputs become scarcer — not where they become abundant. Yet the memo's closing:
> "You are not reading this in June 2028. You are reading it in February 2026. The S&P is near all-time highs. The negative feedback loop has not started... The premium on human intelligence will narrow. As investors, there is still time to assess how much of our portfolio rests on assumptions that cannot survive a decade. As a society, there is still time to act proactively."
> "The canary is still alive."
Takeaways
Why this piece matters
1. Contrarian courage: examining the risk of AI being *too* successful, when consensus focuses on upside. 2. Structured risk analysis: translating vague anxieties into trackable indicators — white-collar unemployment, SaaS ARR growth, high-income ZIP mortgage defaults, labor share of GDP. 3. Timeliness: published after DeepSeek's market shock, amid rapid AI agent progress and peak AI investment enthusiasm.
---
*This is a summary/interpretation of the CitriniResearch report. The original explicitly states it is a thought exercise, not a prediction.*