AI Isn't the Reason American Wages Are Cooling

I keep an eye out for false narratives that take hold of the market, and there is one forming right now around the labor market. The story you are seeing is that American wages are under pressure — and that artificial intelligence is increasingly the brain to blame. One half of that headline is real. The other half is, in my opinion, mostly a story. Get the causal story wrong and you will make the wrong call with your portfolio, so let's separate what the data actually supports.

Start with what is true. Growth in average hourly earnings slowed to 3.1% year over year in August, the least since May 2021. Pay raises genuinely are shrinking, and the direction matches the headline everyone keeps reading.

But "under pressure" only means something next to inflation, and this is where the story starts to fray. Measured against living costs, wages were essentially flat, not collapsing: pay was up 3.5% over the year to July against 3.4% inflation, a real gain of roughly one-tenth of a percent. Workers are not losing purchasing power in the aggregate; the post-pandemic boom in raises is simply fading back toward normal. And the labor market is not unraveling alongside it. August brought 162,000 new jobs, the strongest month since March and roughly three times what forecasters expected, while the unemployment rate held at 4.1%, down from a year earlier. That is the signature of a cooling that is normalizing, not a cooling that is collapsing.

So if the wage deceleration is mostly cycle, what is AI actually doing? The cleanest way to think about it comes from a Federal Reserve Bank of Dallas analysis and the distinction it draws between two kinds of knowledge. AI can replicate codified knowledge — the textbook, routine material in a field. It cannot replicate tacit knowledge — the judgment built from years of experience. That single distinction does the heavy lifting: AI can substitute for entry-level workers, who have the book learning but not the seasoning, while it complements experienced workers, whose tacit knowledge becomes more valuable rather than less.

Watch how that shows up in the data, because it is concentrated, not broad. In the most AI-exposed industries, employment has been flat to down since late 2022 — computer systems design by about 5% — and the hit falls disproportionately on workers under 25, driven by low job-finding rates for new entrants rather than layoffs. Yet wages in those same AI-heavy fields rose faster than the national average. The squeeze is on getting into the field, not on what a field pays once you are in it. AI is compressing the bottom rung of the white-collar career ladder, not flattening wages across the economy.

Two further pieces of evidence point the same way. The Federal Reserve examined job-posting data and found no evidence that firms adopting AI are posting fewer jobs — the post-pandemic slowdown in postings is not driven by AI, and the statistical results are what the researchers call precisely estimated null effects. And the New York Fed notes that fewer than one in ten workers sits in an occupation with meaningful AI exposure, while roughly four in ten have none at all. If AI were broadly crushing wages, we would see it broadly. We do not.

Here is where the investable consequence lands. First, do not run a recession trade off this headline. Cool-but-positive wage growth alongside stable hiring is the backdrop of a soft landing, and the Federal Reserve's next move is being set by inflation, not by a collapsing labor market. Second, the strain that is real is the K-shape: the entry tier — young workers and lower-income households — is the side carrying the pressure, which is a genuine risk to discretionary spending aimed at that cohort, not a reason to fear the whole economy. Third, the demonstrable payoff to AI sits in operating leverage: the firms deploying it at scale appear to be squeezing cost out of routine white-collar labor by not backfilling the bottom of the ladder, which is an earnings tailwind for the operator and an entry penalty for new graduates.

The condition that would overturn this reading is a broadening of the effect. If AI's compression moves up from the entry tier into mid-career, routinized work — a widening of that job-posting gap or a jump in youth unemployment — the narrow story becomes a general one. The Fed itself stresses that this is still an early stage and says it will keep watching. That is the variable to follow, not the fear-of-the-week headline.

Treat the wage story as a reason to check how your portfolio is exposed to the squeezed tier and to the margin story — not as a signal to bolt for the exits.