AI Automates 11.7% of Wages, 5x Visible Impact
MIT's Iceberg Index simulation of 151M US workers across 923 occupations shows AI can already handle tasks worth 11.7% of wages ($1.2T), versus 2.2% ($211B) visibly disrupted—task nibbling leads to job extinction.
Iceberg Index Reveals Task Automation's Hidden Scale
MIT's Project Iceberg simulates 151 million US workers as agents across 923 occupations and 3,000 counties, mapping 32,000 skills to current AI capabilities. It breaks jobs into tasks, tags AI-performable portions, and converts them to wage value: if 30% of a $60k job is automatable, that's $18k exposed. Visible AI adoption disrupts only 2.2% of total wages ($211B/year)—tech layoffs, call centers, junior coding. Hidden exposure in admin, finance, clerical, legal, professional services hits 11.7% ($1.2T), because companies lag deployment due to inertia, budgets, habits. Iceberg Index = hidden/visible ratio (5x), proving current pain is just the tip; agentic AI and browser-use accelerate full evaporation as tasks vanish.
AI already outputs over 1 billion code lines daily, exceeding human volume. Tech (6% workforce) drives 30% S&P value and 1.1 GDP growth points via AI infra spend, but ripples hit non-tech states hardest via secondary collapse: automate office work, kill cleaners, cafes, bodegas.
Challenger, Gray & Christmas data shows 1M+ US layoffs announced for 2025 due to AI, outpacing counts.
PAVM Scores Processes for Targeted Automation
Author's Process Automation Value Model (PAVM) counters Iceberg doom with actionable prioritization: Automation Potential Score (APS) = Complexity + Volume + Automatability + Risk. High APS processes (repetitive, high-volume, low-risk) get robots first for max FTE release, financial benefit. Medium needs simplification; low requires fixing root dysfunction.
From APS derive: Effort Estimate (EE), FTE Release (FR), Financial Benefit (FB), Upskill Index (UI), Net Program Value (NPV). Pair with Reskilling Factory: map freed workers' skills, chart upskill paths, match to high-value roles. Focuses on recycling talent, not sacking—ranks backlog objectively, avoiding gut-feel automation.
Metrics Fail AI's Service Economy Carnage
GDP/unemployment blind to AI: automates services (paperwork, emails, triage) without output change. Hospital admin drops from 6 to 1 hour/patient via AI? Productivity flat, labor vanishes invisibly. GDP counts sales, not savings; tracks new subs but misses $1.2T displacement. Productivity lags until output rises, hiding efficiency. Report correlational, not causal; adaptation slower than change, retraining lags planning cycles.
Build Irreplaceable Skills or Go Manual
AI impersonates soft skills but can't embody: prioritize leadership, communication, creativity, coordination, judgment. Or shift to physical manipulation (hands-on work). Repetitive jobs die regardless of prompt certs—models like Iceberg/PAVM flag high-repetitiveness. Society needs national retraining, tax reforms (dividends, ultra-rich), safety nets; instead, middle class taxed harder amid mortgage/debt traps.