“2028 Global Intelligence Crisis” In-Depth Analysis

Subtitle: When “Intelligence” Is No Longer Scarce, Where Will the Weakest Link in the Financial System Break First?

This article is based on Citrini Research’s scenario projection “THE 2028 GLOBAL INTELLIGENCE CRISIS” (abbreviated as “2028GIC”) published on 2026-02-22. The original emphasizes: “This is a scenario, not a prediction.” Its value lies not in “predicting the future” but in illustrating an underestimated tail risk through a closed-loop chain: If AI becomes too successful, it may not only “boost productivity” but also cause the scarcity assumption of human intelligence to collapse, triggering a reset in financial system pricing and credit structures.

1. What Is This Article Doing: A “Macro Memo from 2028”

The original adopts a “financial history memoir” style: assuming the date as 2028-06-30, using a “quarterly macro memo” tone to review how the crisis evolved from the 2026 “industry shock” into a systemic financial crisis: unemployment at 10.2%, S&P 38% below its October 2026 peak (with hints of deeper declines)—a typical scenario writing approach: framing “the future as already happened” to reduce readers’ cognitive load in understanding complex causal chains, allowing focus on “mechanisms” rather than “predictive numbers.”

Advantages of this approach:

• Reveals the “intermediate process”: financial crises rarely form overnight; they often go through “local issues → risk re-financing → balance sheet linkages → regulatory/liquidity-triggered repricing → systemic collapse.”
• Highlights “fragile points”: even if you disagree with the final outcome, you can see more clearly which markets will distort or break first when a fundamental assumption is shattered.

Its obvious disadvantages:

• Narrative flow ≠ high probability. Scenario projections often assume “accelerated events” at key points, creating dramatic feedback loops.
• Readers may mistake “narrative intensity” for “certainty.”

Therefore, the correct way to read this article is to treat it as a set of “stress test scripts,” then ask yourself:

Which links in this chain are most likely to occur? Which are least likely? If only 30% happen, how will the market price it?

2. Three Core Concepts: Intelligence Premium, Ghost GDP, Friction→0

The most insightful part of the original is how it elevates AI impact from a “productivity tool” to a “scarcity asset.”

2.1 “Intelligence Premium”: Human intelligence as a scarce input

The original states: The modern economy has long assumed a fundamental fact—human intelligence is the scarcest input, thus implicit in wages, asset pricing, and institutional design. It even explicitly says: From labor markets to mortgage markets and tax systems, all are designed based on the premise of “intelligence scarcity.”

If AI makes “analysis, decision-making, creation, persuasion, coordination” replicable and scalable, this premium will be squeezed out, leading to painful re-pricing in the financial system:

• Previously, “high FICO, high income, stable job” loans were risk model cornerstones;
• But if the structural income expectations of high-income jobs are disrupted, the underlying assumptions of risk models will collapse.

2.2 “Ghost GDP”: Productivity rises, but money doesn’t circulate

The original coinages a compelling term: Ghost GDP—“appears in national accounts but doesn’t circulate in the real economy.” It depicts a seemingly contradictory but historically familiar structure:

• Corporate profits temporarily surge due to layoffs and automation;
• Capital (especially “compute power owners”) profits explode;
• Wages stagnate, consumption capacity drops, demand weakens;
• So, “macro data looks strong (productivity, profits),” but people’s actual experience and consumption are poor.

It can be understood as: supply-side efficiency gains + distribution imbalance → insufficient effective demand. This differs from traditional recessions driven by high interest rates suppressing demand: here, the root cause is “the diminishing value of human labor.”

2.3 “Friction to Zero”: Intermediation as a moat is actually friction

The original emphasizes the collapse of “intermediation” layers: over the past 50 years, the US economy built a huge “rent-extracting layer” on “human limitations”—time costs, impatience, information asymmetry, brand familiarity, reluctance to compare prices—all monetized by platforms and intermediaries. When AI agents replace you in searching, comparing, deciding, and executing, the friction that saves money with a few clicks disappears, exposing many business models’ moats as “walls built by friction.”

This logic explains many internet/payment/platform companies: it’s not that the product suddenly worsened, but that users’ “decision costs” suddenly dropped.

3. Scenario Chain Review: From “Industry Shock” to “Systemic Crisis”

Following the original narrative, the core chain is broken into five stages, each dependent on key assumptions.

Stage A (2026): Layoffs initially improve profit margins, market misreads as good news

The starting point is sharp: the first wave of white-collar layoffs caused by AI begins early 2026, but on financial reports “looks good”:

• Labor costs decline → profit margins expand;
• Earnings beat expectations → stock prices rise;
• Companies reinvest profits into AI compute power → stronger AI capabilities.

This is a classic “positive feedback”: short-term financial improvements mask long-term demand issues.
Key assumption: The negative impact of layoffs won’t immediately show up in revenues, and markets will continue to value “productivity narratives.”

Stage B (2027): As agents become widespread, “intermediation rent layer” begins to erode

The original describes AI agent usage becoming standard in 2027: like people using autocomplete without understanding the underlying mechanics. This triggers a chain reaction of “friction to zero” business impacts:

• SaaS (especially process/collaboration/integration) faces competition from “internal tools” that are substitutable;
• Payment networks and card organizations’ interchange fees (2-3%) are bypassed via agents;
• Platforms relying on “users unwilling to bother” are forced into price wars.

Key assumption: Agents can truly perform “end-to-end” across apps/platforms, and regulation/compliance/security won’t be major barriers.
In reality, this may not happen so quickly, but it forces a reassessment: many so-called moats are built on “human laziness to optimize.”

Stage C (2027 Q3): Private credit “software LBO” starts to blow up

This is the most “financialized” and serious part. It provides a clear data point: private credit grew from under $1 trillion in 2015 to over $2.5 trillion in 2026, with much allocated to software and tech deals, especially leveraged buyouts based on “ARR stable deferred cash flows.”
Zendesk is used as a “smoking gun”: when AI agents directly replace “ticket generation—assignment—manual processing,” the “recurring revenue” of such businesses no longer remains “recurring,” as ARR becomes “revenue not yet lost.”

The original insight: markets initially thought this was “manageable” because private credit has lock-in periods, considered “permanent capital,” and unlikely to run into runs. But then, a key reversal occurs:

• Large alternative asset managers treat annuity liabilities of life insurers as capital pools, using them to hold private credit assets;
• When assets are no longer “money good,” regulators increase capital requirements, forcing insurers to recapitalize or sell assets;
• So, the “no run” structure is passively unwound by regulatory and capital constraints.

Key assumption: Credit losses in software/information services are large enough to resonate with insurance capital pools and regulatory capital rules.
This mechanism isn’t just imagination; it aligns with historical patterns where seemingly stable maturity mismatches or regulatory arbitrage cause systemic breaks.

Stage D (2028): From “Loss” to “Admitting Loss”—the crisis tipping point

The original quotes a classic financial history phrase:

The crisis isn’t caused by losses, but by the moment you start recognizing losses.
It then shifts focus to the larger market: the $13 trillion US housing mortgage market.
The key issue is called “The Mortgage Question”: when white-collar income expectations are structurally weakened, are the so-called “prime borrowers” (780 FICO, 20% down, good credit history) still “money good”?
The original emphasizes this differs from 2008:

2008 was “bad loans from the start”;
2028’s scenario is “good loans initially, but the world changed,” and borrowers are taking on debt they no longer believe they can afford in the future.

This creates a second accelerator: as house prices fall, marginal buyers are also income-affected, worsening price discovery, and further depressing wealth effects and consumption.
The chain ultimately pushes stock market retracement toward GFC levels (peak-to-trough 57%).

Key assumption: White-collar income damage is broad and persistent enough to impact mortgage cash flows; house prices and mortgage pressures concentrate in high-tech employment cities and spill over into systemic risk.

Stage E (Policy and Society): Traditional policy tools fail, “tax base = human time” becomes the core contradiction

The original presents a “hard constraint” in policy: government revenue mainly comes from “human time” (wages, salaries, employment). When AI raises productivity but employment/wages decline, fiscal revenue falls below baseline, and society needs more transfers.
This leads to a governance dilemma: “Must give money but can’t collect taxes.”
It even pushes the contradiction into social movements (Occupy Silicon Valley), depicting wealth concentration and social rifts.

4. My Top Three Insights from This Scenario Projection

4.1 It captures that “AI’s impact is primarily a distribution problem, not just productivity”

Many AI narratives focus on “efficiency gains” but overlook “who benefits from efficiency.” The Ghost GDP concept precisely captures a risk: if the gains are concentrated among capital/compute owners rather than labor, macro data and micro experience will diverge, leading to a “high productivity, low demand” structural instability.

4.2 It depicts the “private credit—insurance capital pool—regulatory capital” chain with high realism

Financial crises often originate where “people think there won’t be a run.” Private credit, known for lock-in periods and non-standard assets, was considered less prone to traditional runs; but when linked to insurance funds, offshore reinsurance, and capital requirements, it can trigger “deleveraging” under regulatory and rating triggers.
This chain feels very real.

4.3 It reminds you that the core of the mortgage market isn’t “can you pay now,” but “are future income expectations stable”

The most damaging aspect isn’t “bad debts already occurred,” but “the assumption of high-quality borrowers’ income being structurally shaken.” Even if short-term payments are maintained via savings, HELOC, or 401(k) withdrawals, consumption will decline first (as they cut discretionary spending).
This aligns with many recession experiences where “consumption leads.”

5. My Top Three Weak Points and Over-Discounted Aspects

5.1 “Friction to Zero” is portrayed too quickly: real-world frictions aren’t just from regulation

In reality, many frictions come from regulation, compliance, KYC, data silos, responsibility, fraud risks, offline fulfillment, etc. Even if agents are smart, they need cross-organizational authorization and integration.
Thus, “platform moats will weaken” may be true, but “zero within a year” is likely exaggerated.

5.2 The speed of “white-collar replacement” is doubtful: more likely initial “entry-level job collapse + experience premium rise”

Research by Dallas Fed economist J. Scott Davis (2026-02-24) offers a more data-driven view: AI may both replace and assist labor—more easily replacing codifiable “book knowledge tasks,” but more likely augmenting experience-based work relying on tacit knowledge; data shows high exposure industries see more employment decline among under-25s.
This suggests the short-term impact may be narrower: narrower entry points for new graduates, broken career ladders, rather than all white-collars losing jobs in 24 months.

5.3 “Agents using crypto for settlement” is more a narrative enhancer than a necessity

The original mentions agents potentially bypassing traditional payment networks for cheaper settlement. But this isn’t essential for the crisis chain; the key is the compression of “intermediation rent layers” and profit revaluation.
So, readers shouldn’t see “migration to crypto payments” as the main indicator of the scenario’s validity.

6. Turning Scenario Projections into Actionable Monitoring Dashboards: What Should We Watch?

Scenario value lies in breaking it into indicators. Here’s a practical monitoring checklist (not prediction-based, just observation-based):

Monitoring Targets

Example Indicators

Why It’s Key

Which part of scenario it triggers

White-collar employment structure

AI-exposed industry employment/salaries, under-25 job acquisition rate, hiring freeze duration

Verify early signs of “entry point collapse/distributional income damage”

Stages A / D

Consumption and Credit

Credit card balances in high-income cities, HELOC draws, 401(k) early withdrawals, discretionary spending

Verify “mortgages still being paid, consumption declines first”

Stage D

Software/Consulting Boom

Software subscription renewal rates, ARR pressure, info service profit margins and orders

Verify “industry shock → credit issues”

Stages B / C

Private Credit Stress

Secondary loan prices, default rates, restructuring volume, LP redemption pressure

Verify “software LBO blow-up spilling over”

Stage C

Insurance Regulation/Capital

NAIC/state regulator changes in capital requirements for private-rated assets, insurer rating outlooks

Verify “passive deleveraging of ‘permanent capital’”

Stage C

Regional Mortgage Pressure

Early delinquencies and house price index changes in high-tech/financial cities

Verify “re-pricing of prime mortgages”

Stage D

Macroeconomic Divergence

Disparity between productivity/profits and actual wages/consumption

Verify Ghost GDP formation

Stages A / E

Core idea of this panel:

Don’t guess whether AI will “suddenly reach AGI,” but observe early signs of “distribution and credit” issues as described in the original.

7. Personal, Corporate, and Investment Implications: Three “New Frameworks”

7.1 For Individuals: Position yourself in “high tacit knowledge, high experience premium” zones

If Dallas Fed’s conclusions are closer to reality, the real danger is “career ladder disruption”: entry-level jobs replaced by AI, making it hard for young people to gain experience.
So, individual strategies should be:

Choose fields that build tacit knowledge and judgment (require on-site presence, responsibility, coordination);
Use AI as an amplifier: accelerate information processing, dedicate more time to “decision quality, cross-team collaboration, understanding complex systems.”

7.2 For Companies: Moats should shift from “friction” to “trust, data, fulfillment, ecosystem”

As agents lower search and comparison costs, moats based on “channels/traffic/information gaps” weaken. Companies should focus on:

Verifiable quality and fulfillment capabilities (offline/supply chain/service systems);
Compliance and accountability (risk-bearing, liability);
Unique data and ecosystem synergy (not just info, but closed-loop processes).

7.3 For Investors: Beware of “leverage assets based on stable future assumptions”

The most painful insight: crises don’t start with “the most fragile,” but with “the most trusted” (prime mortgages, ARR loans, permanent capital).
Investment takeaways:

In high-uncertainty AI periods, assets priced on “future stability” (high leverage, long durations, perpetual growth/stable renewal) need higher margins of safety;
Focus on balance sheet and cash flow resilience against demand drops;
Don’t trust that “no runs structurally,” as regulation and capital rules can create equivalent runs.

8. Conclusion: This Isn’t “Doomsday Prophecy,” But a Stress Test Script

The most valuable aspect of “2028GIC” is that it forces you to admit:

If AI truly makes “human intelligence no longer scarce,” many of our current systems and financial pricing frameworks are incompatible.

But you must also acknowledge: real-world feedback loops won’t unfold exactly as scripted; frictions, regulation, social adaptation, and new job creation will alter the path.
So, the most mature approach is to:

Retain its mechanistic insights (distribution, credit, institutional vulnerabilities),
Reduce certainty about timing (don’t bet on full realization within 24 months),
Use indicators to track whether it is “partially materializing.”

When scenario projections can be broken into monitoring dashboards, they transform from “stories” into “tools.”

References

Citrini Research & Alap Shah: “THE 2028 GLOBAL INTELLIGENCE CRISIS” (Feb 22, 2026)
Federal Reserve Bank of Dallas: J. Scott Davis, “AI is simultaneously aiding and replacing workers, wage data suggest” (Feb 24, 2026)