# Torsten Slok Apollo AI wage study: pay, not jobs _Apollo's Torsten Slok finds AI exposure since ChatGPT has lowered wage growth but not employment across 300 occupations._ **Published:** 2026-09-05 **Source:** https://www.startuphub.ai/ai-news/market-research/2026/torsten-slok-apollo-ai-wage-study-pay-not-jobs --- AI is squeezing paychecks before it cuts headcount. The Torsten Slok Apollo AI wage study, detailed on [Bloomberg Podcast](https://www.youtube.com/watch?v=zDOs7iuF4do)’s Wall Street Week, finds highly exposed occupations are already seeing weaker wage growth since ChatGPT arrived. Apollo study on AICore Torsten Slok compares 300 occupations before and after ChatGPT using real dataFrom the article 2 mentionsThe Torsten Slok Apollo AI wage study, detailed on Bloomberg Podcast’s Wall Street Week, finds highly exposed occupations are already seeing weaker wage growth since ChatGPT arrived.Wages take the hitOutcomeHighly exposed occupations show weaker wage growth since ChatGPT launchedChatGPT arrivalDriverNatural experiment split by pre and post ChatGPT wage and employment trendsFrom the article 3 mentionsHe compared hundreds of occupations before and after ChatGPT and split them by AI exposure.Pay squeezed firstEffectAI compresses compensation before any reduction in headcount shows upFrom the articleHigh-attrition, repetitive roles like customer service and paralegal work are automated first and backfilled at lower rates, which drags pay even if payrolls hold.Apollo study on AICoreTorsten Slok compares 300 occupations before and after ChatGPT using real dataFrom the article 2 mentionsThe Torsten Slok Apollo AI wage study, detailed on Bloomberg Podcast’s Wall Street Week, finds highly exposed occupations are already seeing weaker wage growth since ChatGPT arrived.tracksChatGPT arrivalDriverNatural experiment split by pre and post ChatGPT wage and employment trendsFrom the article 3 mentionsHe compared hundreds of occupations before and after ChatGPT and split them by AI exposure.measuresMeasure exposure levelsContextOccupations grouped as above or below 50 percent AI exposure using usage dataWages take the hitOutcomeHighly exposed occupations show weaker wage growth since ChatGPT launchedJobs stay steadyOutcomeEmployment levels remain largely unchanged across AI-exposed occupations overallshowsPay squeezed firstEffectAI compresses compensation before any reduction in headcount shows upFrom the articleHigh-attrition, repetitive roles like customer service and paralegal work are automated first and backfilled at lower rates, which drags pay even if payrolls hold.drivesPolicy and profit impactEffectFindings carry major implications for labor policy and corporate profit margins Apollo chief economist Torsten Slok co-authored the first study of its kind to use actual employment and wage data, not theory. He compared hundreds of occupations before and after [ChatGPT](https://www.startuphub.ai/ai-news/artificial-intelligence/2026/openai-s-chatgpt-admin-plugin-streamlines-workspace-management) and split them by AI exposure. The full discussion can be found on **Bloomberg Podcast**'s YouTube channel. ![](https://img.youtube.com/vi/zDOs7iuF4do/maxresdefault.jpg) Fan Favorite: AI Hits Wages, USMCA Under Pressure, Colorado River Crisis | Wall Street Week, from Bloomberg Podcast ## What the study actually measured The team categorized about 300 occupations by how exposed they are to AI using actual usage data. They then ran a natural experiment comparing wage and employment trends before ChatGPT with trends after. Occupations more than 50 percent exposed to AI were grouped against those less exposed. That cutoff was chosen to separate high and low exposure cleanly. ## Are AI-exposed workers losing their jobs? No, not yet at scale. Slok said the employment effect across all groups is insignificant so far. The displacement debate has focused on layoffs, but the data shows automation has not translated into aggregate job losses. Instead, weekly data from the Centers for Business Formation shows US business creation at the highest level ever, as AI tools make it easier to start new firms. ## Why wage growth is taking the hit High-exposure occupations saw measurably weaker wage growth than low-exposure ones. Slok pointed to knowledge work where AI substitutes for tasks without eliminating the role outright. Boston Consulting Group senior adviser Diane Gerson, former IBM head of HR, described the same mechanism on the ground. High-attrition, repetitive roles like customer service and paralegal work are automated first and backfilled at lower rates, which drags pay even if payrolls hold. ## What this means for policy and profits For now, policy makers should wait and see. Slok said Washington does not know whether AI will push unemployment to 10 to 15 percent or push growth and hiring higher, so the toolkit depends on which scenario emerges. Corporate margins tell a similar wait-and-see story. Even as enterprises pour money into AI, including the [nvidia](https://www.startuphub.ai/ai-news/artificial-intelligence/2026/nvidia-s-huang-ai-job-fears-are-nonsense) $500 billion investment discussed on the show, margins for the S&P 493 outside the Magnificent 7 have not yet risen. Slok remains optimistic. He called AI a miracle drug for productivity and employment if adoption follows the S-curve, though he warned effects may grow over time and pressure on the [Trump administration](https://www.startuphub.ai/ai-news/technology/2026/h-1b-visa-fee-struck-down-ai-in-law-explained) to choose between cooling an overheating economy or supporting displaced workers is not yet decided. --- Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory.