When AI Takes the Job, Who Misses the Mortgage Payment?

The Question Nobody Is Asking

The conversation about AI and jobs has produced two camps. Optimists cite historical precedent — technology creates more jobs than it destroys, eventually. Pessimists cite the pace — METR time horizons are doubling every 4.3 months, and frontier models can now handle 14-hour professional tasks autonomously.

Both camps are debating employment. Neither is asking the downstream question: when a worker displaced by AI misses their mortgage payment three months later, where does that show up in the data? When they stop making car payments six months after that, which indicators move? When they drain their savings and raid their 401(k) to stay current, does anyone connect the capability curve to the hardship withdrawal rate?

The answer, right now, is no. No federal agency, no research institution, and no financial publication is systematically tracking the pipeline from AI capability growth through workforce displacement to household financial distress. The American Distress Index and the American Worker Index exist to close that gap.

The Upstream Pipeline

The pipeline from AI deployment to household default runs through four stages. Each has measurable data. None are being tracked as a connected sequence — until now.

Stage 1: Capability

AI autonomous task horizons — measured by METR as the duration of tasks frontier models can complete at 50% success — reached 719 hours (4.3 weeks) in February 2026. In early 2023, that number was 4 hours. The doubling time: 4.2 months (95% CI: 105-155 days).

At 4 hours, AI replaced junior analyst tasks. At 60 hours, it could handle multi-day professional workflows — code review, document production, research synthesis. At 350 hours, it handles what used to be two weeks of mid-career knowledge work. At 719 hours, it manages month-long autonomous projects.

The 80% reliability horizon — the task length where models succeed four times out of five — also moved, from roughly 30 minutes to over an hour. Reliability, not peak capability, is what drives enterprise adoption decisions.

Stage 2: Adoption

The U.S. Census Bureau’s Business Trends and Outlook Survey (BTOS) measures how many businesses are actually using AI. In September 2023, 3.7% of employer businesses used AI in production of goods and services. By September 2025, that number reached 10.0% — nearly a threefold increase in two years.

When BTOS expanded its question in November 2025 to include any business function (not just production), the number jumped to 17.3%. The Information sector led at roughly 40%. Professional, scientific, and technical services — the sector that employs the most knowledge workers — expected a 3 percentage point increase in the following six months.

These are not pilot programs. These are production deployments at businesses that employ people.

Stage 3: Displacement

Challenger, Gray & Christmas began tracking AI-attributed layoffs in 2023. In 2025, companies cited AI as the reason for approximately 140,000 job cuts — a category that did not exist three years earlier. January 2026 added another 7,600.

That 140,000 figure represents about 4.5% of total reported layoffs in 2025. It is almost certainly an undercount. Challenger tracks announcements, not quiet non-replacements of departing workers. When a company lets attrition reduce headcount because AI handles the workload, that displacement never appears in any layoff report.

Total announced job cuts hit 108,400 in January 2026 — the highest January since the Great Recession. The 2025 full-year total reached 1.2 million, up 58% from 2024. AI is not the only driver — DOGE-related government cuts accounted for the largest share — but it is the fastest-growing category.

Stage 4: The Gap

Here is where the analysis breaks down everywhere except at American Default. After a worker loses their job — for any reason — a predictable sequence unfolds:

1. Month 1-3: Savings buffer absorbs the shock. If there is a buffer.
2. Month 3-6: Credit cards and hardship withdrawals bridge the gap. Debt service rises.
3. Month 6-12: Missed payments appear in delinquency data. Auto loans go first (can’t lose the car), then credit cards, then mortgage.
4. Month 12+: Charge-offs, collections, foreclosure filings.

This 3-12 month lag is why AI displacement does not yet appear in the ADI composite. The capability acceleration is real. The adoption is real. The layoffs are real. But bank-reported delinquency and charge-off data — the foundation of the ADI — operates on a quarterly lag. The pipeline is forming. The downstream data hasn’t caught up yet.

The Demographic Overlap Nobody Has Named

This is the connection that Brookings, McKinsey, and the academic AI-and-labor literature have not made.

The occupations most exposed to AI displacement — clerical support, customer service, data entry, bookkeeping, administrative assistance, basic legal and financial processing — are not randomly distributed across the income spectrum. They are concentrated in the $30,000-$55,000 household income range. They are disproportionately held by workers without four-year degrees. They are disproportionately held by women, by Black and Hispanic workers, and by workers in their 20s and 30s.

Now look at who holds FHA mortgages: first-time buyers with credit scores as low as 580, minimum 3.5% down payments, median income in the low-to-mid $50,000s, and effectively zero financial cushion beyond the down payment itself. FHA delinquency is already at 11.52% — 6.5 times the conventional rate of 1.78%.

Look at who carries the subprime credit card balances concentrated at smaller banks, where delinquency runs 6.62% compared to 2.84% at the top 100. Look at who is driving auto loan delinquency to 5.21%, its highest level since 2010.

It is substantially the same population.

The workers most likely to be displaced by AI are the borrowers most likely to default when their income disappears. This is not a coincidence. It is the same underlying vulnerability — limited education-based bargaining power, thin savings buffers, high debt-to-income ratios — expressing itself on both sides of the same economic equation.

The K-Shape Gets Another Dimension

The Two-Economy Problem documented a K-shaped divergence in household finance: prime borrowers doing fine, subprime borrowers deteriorating across every category. AI displacement adds a third axis to that divergence.

Consider three workers:

Worker A is a software architect earning $180,000. AI tools make them more productive. Their company grows. Their job is secure, possibly enhanced. They have 8 months of savings, a conventional mortgage at 3.2%, and a 780 credit score.

Worker B is a customer service representative earning $38,000. Their employer deploys an AI system that handles 60% of inbound calls. The team is cut from 200 to 80. Worker B is let go. They have 3 weeks of savings, an FHA mortgage at 6.8% with 3.5% equity, a $4,200 credit card balance at a regional bank, and a $22,000 auto loan.

Worker C is a bookkeeper earning $45,000 at a mid-size firm. AI accounting software makes 80% of their work unnecessary. Their position is eliminated. They have $1,100 in savings, no mortgage (they rent), $8,500 in credit card debt across two cards, and a $19,000 auto loan.

Workers B and C enter the default pipeline. Their savings last weeks, not months. Their debt service ratio was already stretched. Within 90 days they are behind on the car payment. Within 180 days they appear in delinquency statistics. Worker B’s FHA mortgage enters the 11.52% — and climbing. If Worker B acts early, loss mitigation options exist — but only if they call their servicer before the 120-day mark.

Worker A’s experience does not offset Workers B and C in any meaningful way, but their continued performance pulls every national average down. The blended data looks fine. The disaggregated data tells a different story.

What the Data Says Right Now

The American Worker Index currently reads 83.0 — Crisis. The five AWI components:

The American Distress Index reads 57.1 — Elevated. The ADI’s Labor Market component is currently suppressing the composite — initial unemployment claims remain historically low, dragging that component’s Z-score to -0.45. This means the traditional labor market data does not yet reflect AI-specific displacement at a level that moves the needle.

But the Buffer Depletion component — the ADI’s largest at 30% weight — is at Z = +0.57 and rising. The personal savings rate is 3.6%. Hardship withdrawals are at a record 6.0%. Households are already burning through the buffers they would need to survive a job loss.

The 9-quarter lag between Buffer Depletion and Debt Stress (r = 0.69) means the buffers being consumed today are the delinquencies of 2027-2028. If AI displacement accelerates into that buffer-depleted environment, the impact on default data will be amplified — not because AI displacement is inherently different from other job loss, but because it arrives when households have no remaining cushion.

The Measurement Gap

No existing framework connects these four stages into a single tracked pipeline:

• Brookings publishes an AI Exposure Index ranking occupations by susceptibility to automation. It does not track what happens to those workers’ financial obligations after displacement.
• McKinsey Global Institute estimates that 30% of work hours could be automated by 2030. It does not connect this to delinquency or default data.
• The Federal Reserve tracks employment, wages, and financial distress — but does not disaggregate any of these by AI exposure.
• Challenger tracks AI-attributed layoffs but not the downstream financial outcomes of displaced workers.
• Census BTOS tracks business AI adoption but not the household financial impact of that adoption.

Each institution measures one stage of the pipeline. Nobody measures the full sequence. The AWI was built to fill that gap on the displacement side. The ADI tracks the downstream consequences. Together, they are the only framework that connects AI capability growth to household financial distress through observable, quarterly-updated federal data.

What to Watch

Five data points will determine whether the AI-to-default pipeline activates at scale:

1. AI-attributed layoffs crossing 20,000/month sustained. The current monthly average is roughly 12,000. At 20,000+, the volume becomes large enough to affect localized default rates in tech-heavy metros.

2. BTOS adoption rate crossing 20% in production. Currently 10%. When 1 in 5 businesses is using AI in production, the displacement effects become structural rather than anecdotal.

3. Initial unemployment claims rising in NAICS 51, 52, and 54. Information, Finance/Insurance, and Professional/Technical services are the sectors where AI deployment is heaviest. Sector-specific claims data — available from BLS — would be the earliest signal of AI-driven layoffs large enough to matter.

4. FHA delinquency continuing to rise while conventional holds flat. This pattern — already at 6.5x — would confirm that displacement-driven distress is concentrating in the same low-income, low-savings population most exposed to AI.

5. Savings rate falling below 3.0%. At 3.6%, the buffer is thin. Below 3.0%, any significant shock — including accelerating AI displacement — hits households with effectively no financial cushion. The last time savings fell below 3.0% was the 2005-2007 pre-crisis period.

None of these are predictions. They are observable checkpoints. If AI displacement drives household default at scale, these are the numbers that will move first. We will document them as they happen.

The Bridge

The American Distress Index was built to track household financial distress from all causes. The American Worker Index was built to track the specific upstream pressure of AI-driven workforce displacement. The connection between them — the bridge from capability to adoption to displacement to default — is the question this project exists to answer.

The AWI reads 83.0. The ADI reads 57.1. The gap between those two numbers is the lag between displacement and default. It is the time it takes for a lost job to become a missed payment, for a missed payment to become a delinquency, for a delinquency to become a charge-off.

That gap will close. The question is how fast and how far.

We don’t predict. We track.

Data sources: METR Time Horizons (Model Evaluation & Threat Research), Census Business Trends and Outlook Survey (U.S. Census Bureau), Job Cut Announcements (Challenger, Gray & Christmas), FHA Delinquency Rate (MBA National Delinquency Survey), Personal Saving Rate (BEA via FRED), Vanguard How America Saves. For unemployment and labor market trends, see our Unemployment & Job Market Statistics roundup. Full indicator catalog at americandefault.org/indicators.