Deep Dive
The Study's Sample Problem
Kevin starts by acknowledging the RAND study showing 21,559 firms with high AI adoption saw 12% employment growth at entry level and 10% overall. But he immediately flags the methodology: the study only includes companies using Ramp, a fintech platform, which skews the sample toward startups and tech-forward businesses, not representative of the broader American economy. More critically, the study only counted firms that remained in business throughout the entire period and maintained spending above the minimum threshold. This means any company that went bankrupt, pivoted away from AI, or downsized below five employees simply vanished from the dataset. Kevin's point: you're not seeing job losses from AI failures because those failures were literally removed from the analysis.
Tech Firms Are the Only Winners
Kevin digs into page 30 of the study, where RAND buried the real finding: job gains were concentrated almost entirely in information technology, software, internet, and media companies. For every other sector — retail, manufacturing, services, finance outside of fintech — there was no statistically significant increase in hiring. This destroys the narrative that AI broadly creates jobs across the economy. Kevin notes that even the gains in tech might not be real productivity wins. Many of these firms are venture-backed startups that hired aggressively because they had fresh capital, not because AI made them more efficient. He's explicit: companies using a $33-per-employee-per-month AI subscription as their qualifying 'high adoption' spend are probably just adding headcount as a vanity metric to show investors that their capital is being deployed.
The Missing Long-Term Picture
The study measured job growth during the workflow development phase, roughly 6 to 12 months after AI adoption. But the researchers themselves acknowledged lower confidence in the long-term estimates, which is a massive red flag. Kevin explains: companies are hiring during the experimentation phase, but the study didn't track what happens after two years when AI adoption typically shifts from 'building workflows' to 'optimizing efficiency.' That's when layoffs arrive. He uses his own company, House Hack (now Reinvest), as a case study. When they deployed AI tools, they cut their G&A staff from 12 to 3 people — firing nine employees — while simultaneously hiring software developers to build proprietary AI products. The net effect looked like growth, but it masked real job destruction in administrative roles. Meanwhile, Goldman Sachs data shows AI-affected sectors are seeing net job losses of 10,000 to 12,000 jobs per month, which the RAND study completely misses.
Why This Matters for Investors
Kevin's core argument is that the study conflates temporary hiring during AI buildout with genuine job creation. Early spending on AI tools, hiring to experiment with workflows, and venture capital fueling expansion are not the same as sustainable productivity gains. If AI truly made workers more efficient, companies would be keeping smaller teams and paying them more for output. Instead, Kevin observes that many firms are taking venture money and converting it into salaries without investing in real assets — hardware depreciates, startups can fail, and when the CapEx cycle slows or venture funding dries up, those jobs vanish. The contractors and employees hired to build out data centers are especially vulnerable. Kevin's message: don't get complacent because David Sacks tweeted a headline. The actual data says AI hiring is concentrated in venture-backed tech firms during the workflow phase, not a broad economic boom. Once efficiency demands kick in and the innovation cycle matures, expect significant job losses.