Every AI Jobs Program Skips the Same Step

On June 25, two governments made opposite bets on AI's labor impact. The federal government launched RAISE US, a retraining fund with more than $500 million raised toward a $1 billion goal for education, wage insurance, and short-time compensation in four states ^[1]. California launched a dashboard linking unemployment claims to AI-exposure measures by occupation and demographics ^[2]. One built the response. The other built the instrument to see who needs it. That same divide runs through every surveyed government. Singapore approved a million AI-Ready SG training places while a member of parliament was still calling for the government to build localized market intelligence to guide workers, the measurement tool that hasn't been built yet ^[3]. China has defined 20-plus new AI professional roles with training subsidies, deciding where workers should go without tracking where they're displaced from ^[4]. Canada, Thailand, the UK: each has a training, welfare, or skills program touched by AI, none with an AI-specific displacement instrument ^[5][6][7][8]. The private sector has been measuring what governments won't. PwC's Global AI Jobs Barometer analyzes more than a billion job ads across 27 countries. Challenger, Gray & Christmas has tracked 87,714 AI-cited layoffs in the first five months of 2026, more than 2024 and 2025 combined. Harvard Business School used the Labor Department's own O*NET database to find a 13% drop in postings for repetitive-task roles since 2022. The OpenAI Foundation's $250 million commitment names measurement as its first pillar ^{[9][10][11][12]}. None of these instruments can trigger a government retraining deployment. The actors who can measure can't respond. The actors who can respond haven't measured. The problem is not that no data exists. The BLS produces occupational employment data. The Bank of Canada produces employment data. Both conclude there is no widespread AI displacement. The Bank compares AI to computer adoption while simultaneously noting weak entry-level hiring in coding and customer service ^[5]. The former BLS commissioner reads the same apparatus and reaches the same conclusion:

It could be. It likely will be disruptive, but the data is telling us right now that disruption is not yet here, and that we have time to plan. — Erika McEntarfer

Yet Stanford's Erik Brynjolfsson finds a 16% decline in entry-level jobs in AI-exposed occupations ^[13]. The contradiction is the measurement gap in plain view. Government instruments are designed to see aggregate employment, which remains stable. The displacement pattern is tier-specific: entry-level erosion invisible in the aggregate. Bloomberg economist Eliza Winger names the problem precisely:

hiring remained selective and primarily focused on critical roles or attrition replacement. — Federal Reserve System

Tech companies announced 38,242 cuts in May 2026, the highest monthly total in nearly two years, up 65% year over year ^[14]. Aggregate data says calm. Occupational data says churn. No government instrument was designed to reconcile the two until California's. Newsom's May 21 executive order mandated a two-step architecture: a 90-day research review, then a dashboard to track AI's sectoral impacts, with recommendations to update the WARN Act within 180 days ^[15]. The order came as Trump signed a directive preventing states from regulating AI. Newsom's position:

California has never sat back and watched as the future happened to us – and we won’t start now. — Gavin Newsom

The tracker that launched June 25 is that dashboard. Its initial data shows a sustained increase in claims among college-educated workers in high-exposure occupations since 2022, concentrated in the Bay Area ^[2]. Officials say they intend to use the monthly data to deploy targeted retraining, career counseling, and health coverage ^[2]. The instrument is built. The loop is declared. It has not closed. No retraining program has been triggered by the tracker's data. Researchers caveat that overlaying AI-exposure measures onto unemployment claims establishes correlation, not causation ^[2]. And California's own labor leaders argue the deliberate pace may be too slow for what it's measuring. Lorena Gonzalez of the California Labor Federation:

Catastrophic job loss from AI is not inevitable, it’s a political choice. — Lorena Gonzalez

She frames catastrophic job loss as a political choice, not an inevitability ^[15]. The critique is not that the tracker is wrong but that the measure-then-respond loop may close too slowly to catch what it was built to see. PwC's Pete Brown describes what's already disappearing:

AI is removing some of the routine work that once acted as an apprenticeship, while increasing demand for judgement, leadership and adaptability much earlier in careers. — Pete Brown

California has built the only government instrument designed to see that erosion in real time and respond to it. Whether the response can keep up, or whether the careful architecture of measure-then-respond is already too deliberative for the pace of displacement, is the question the tracker's first months will answer.