AI labor market data point to pay squeeze, not layoffs yet

AI labor market research from Apollo finds early effects in slower wage growth rather than broad job cuts. The gap is widest in exposed roles.

Torsten Slok, Apollo Global Management Inc.’s chief economist, said work with co-author Sania Edlich found a clearer pay effect than an employment shock. Their review of roughly 300 occupations found wages in high-exposure jobs grew 6.7% more slowly than in low-exposure jobs after ChatGPT broadened corporate use of generative AI.

Apollo splits 300 occupations

The study divided occupations into groups with higher and lower exposure to AI, then compared performance before and after ChatGPT became widely available. That design uses the same labor market period for both groups, making the difference between them the central measure.

Slok said the employment effect so far has been “insignificant,” while the wage signal is easier to see. The finding does not mean AI has had no effect on work; it indicates that early pressure may be showing up in pay growth before payroll counts move.

Payroll data offer a check

Other evidence is less benign. The Bureau of Labor Statistics reported that 18 AI-exposed occupations, covering about 10 million jobs, recorded a 0.2% employment decline in the 12 months through May 2025, while total payrolls increased 0.8% over the same period.

Goldman Sachs Group Inc. economists said in May that job openings fell faster in fields “highly exposed to AI substitution” than in less exposed areas. That points to weaker hiring demand even where companies have not announced large job cuts.

Attrition masks some reductions

The Apollo finding was heavier among lower-paid workers, according to Slok. That matters because many exposed roles include repeatable administrative or customer-facing tasks where employers can slow wage increases or reduce hiring without immediately cutting existing staff.

Diane Gherson, the former chief human resources officer at International Business Machines Corp., said some reductions may be hard to detect in headline layoff data. She said companies can shrink teams in high-turnover, lower-wage areas such as customer service by replacing fewer workers who leave.

Startup gains complicate the picture

Slok also said AI has accelerated business formation, with new company creation now at a record high. “So far the dominating effect has been that there is also a much more dynamic economy where people can now invent ideas, use agents, use loops, graphs to come up with ideas and as a result, create more businesses,” he said.

That argument gives the labor-market debate two channels rather than one. AI can restrain wage growth in exposed jobs while also lowering the cost of starting firms, which may create demand elsewhere if those businesses survive and expand.

Three paths for AI hiring

If Apollo’s wage-first pattern holds, the macro effect would be a cooler pay channel for investors tracking US inflation, without an immediate jump in unemployment. For Apollo, that would support Slok’s view that AI is changing bargaining power before job totals; for software and services companies, it would favor tools sold as productivity aids rather than headcount replacements.

If the BLS and Goldman signals broaden, the macro risk shifts toward weaker hiring and slower household income growth. That outcome would challenge Apollo’s limited-displacement reading and put more pressure on employers in exposed occupations, especially customer service, administrative support and other roles where attrition can lower staffing quietly.

If AI-driven business formation keeps rising, the macro effect would depend on whether new firms add durable jobs faster than exposed roles shed demand. Apollo’s thesis would then rest on a balance between wage restraint and startup creation, while the wider technology sector would face pressure to prove that AI tools create revenue growth, not only cost savings.