→Claude Just Joined Your Meetings, But Only on Google Meet →An AI Vendor Breach Just Leaked Uber Driver IDs and Notion Chat Logs at the Same Time →A Safety Reminder Cut Harmful AI Medical Advice by a Third. It Still Got 1 in 10 Wrong. →A Power Bank Just Launched That Costs 75 Percent More to Not Catch Fire →Firefox's New Mini Window Is Genuinely Useful, With One Annoying Catch →How to Actually Check if Your Email or Password Was Leaked in a Data Breach →Google's New AI Model Reads DNA Cell by Cell, and the Numbers Are Real →Perplexity's Comet Browser Is Free Now: What You Gain and What You're Trusting It With →Dropbox Just Spent $350 Million on a Storage Startup to Escape Being Dropbox →AI Is Not Taking Your Job. It Might Just Be Freezing Your Raise →NotebookLM Is Gone, Gemini Notebook Is Here: What Actually Changed →Four AI Models, Five Accents, One Hierarchy: What a New Study Found →Claude Just Joined Your Meetings, But Only on Google Meet →An AI Vendor Breach Just Leaked Uber Driver IDs and Notion Chat Logs at the Same Time →A Safety Reminder Cut Harmful AI Medical Advice by a Third. It Still Got 1 in 10 Wrong. →A Power Bank Just Launched That Costs 75 Percent More to Not Catch Fire →Firefox's New Mini Window Is Genuinely Useful, With One Annoying Catch →How to Actually Check if Your Email or Password Was Leaked in a Data Breach →Google's New AI Model Reads DNA Cell by Cell, and the Numbers Are Real →Perplexity's Comet Browser Is Free Now: What You Gain and What You're Trusting It With →Dropbox Just Spent $350 Million on a Storage Startup to Escape Being Dropbox →AI Is Not Taking Your Job. It Might Just Be Freezing Your Raise →NotebookLM Is Gone, Gemini Notebook Is Here: What Actually Changed →Four AI Models, Five Accents, One Hierarchy: What a New Study Found
Panda Tech Bytes
All Posts
Latest AI Trends Guides Tools Tech Pulse Hardware & Gear Funding Watch Risk & Reality What It Means for You Security Watch On Your Desk
AI Is Not Taking Your Job. It Might Just Be Freezing Your Raise

AI Is Not Taking Your Job. It Might Just Be Freezing Your Raise

A new analysis from Apollo Global Management finds that workers in AI-exposed jobs are not losing their positions, but their pay raises have slowed sharply, costing millions of people thousands of dollars a year in wage growth they never got.

For the last few years, the AI jobs conversation has mostly been about one question: will this replace me? A new analysis from Apollo Global Management, the asset manager behind one of Wall Street's most closely watched economic research desks, suggests the more immediate risk is quieter than that. Workers in jobs most exposed to AI are not losing their positions in large numbers. They are just not getting raises the way everyone else is.

What the study actually measured

Apollo's chief economist Torsten Slok and a co-author matched 321 occupations to a decade of United States labor statistics, from 2015 through 2025, and layered in AI usage data from the Anthropic Economic Index, which tracks how people actually use AI models rather than relying on theoretical guesses about which jobs "could" be automated. That is a meaningfully different approach than most AI jobs studies, which tend to estimate exposure based on task descriptions rather than observed behavior.

The headline result: wages in highly AI-exposed occupations grew 6.7 percentage points more slowly after 2023 than wages in low-exposure occupations. Employment levels in those same high-exposure jobs showed no statistically significant change. In plain terms, people are not being let go. Their pay is just falling behind.

The dollar figure behind the headline

Apollo's researchers translated that wage gap into a rough aggregate cost: about $28 billion a year in foregone wage growth, spread across an estimated 5.8 million workers, or close to $4,800 per person annually. That is not money taken out of anyone's current paycheck. It is the raise that did not happen, compounding year over year.

The effect was not evenly distributed. The study broke results out by income quartile:

Service workers in AI-exposed roles showed the steepest gap of all, at 24.3%, though Apollo flagged that figure as based on a small sample and urged caution in reading too much into it.

Why wages and not jobs

The explanation researchers offer is fairly intuitive once you sit with it. Laying someone off is disruptive, expensive, and visible. Quietly slowing how fast their pay grows is none of those things. If AI tools let a customer service rep, a financial analyst, or a computer programmer handle a bigger workload without a bigger raise, a company captures the productivity gain without the headline risk of a layoff announcement. Slok described it as AI "hitting paychecks rather than payrolls."

That lines up with a separate, unrelated finding buried in the same labor data: AI-exposed occupations as a group still grew payrolls, just not as fast as other sectors, while wage growth in those same jobs fell well behind. The jobs are there. The raises are not keeping pace.

The honest caveats

This is where it is worth slowing down, because the study itself is careful about its own limits, and some outside labor economists have pushed back on how far to take it.

Apollo could only match 321 of roughly 800 Bureau of Labor Statistics occupation codes to usable AI exposure data, and only 11 occupations cleared the study's bar for "high exposure." The underlying usage data comes from a single AI company's product, not the whole market. And the dramatic service-worker number, the one most likely to get quoted out of context, is explicitly flagged by the authors as resting on a small sample. Apollo itself describes the findings as early evidence, not a settled conclusion, and some labor researchers have said the sample is too thin to be fully convincing on its own.

None of that erases the core finding. It just means the honest version of this story is "a well constructed but still early analysis found a real wage gap," not "AI is definitively suppressing millions of paychecks."

What this means if you work in an AI-exposed job

If your role touches writing, analysis, customer support, coding, or anything else that generative AI tools handle reasonably well today, this study is a useful nudge to look past the "will I get replaced" fear and ask a more immediate question: is my pay keeping pace with how much more I am now expected to produce. Several separate surveys this year have found that employees report higher productivity expectations once their employer rolls out AI tools, without a matching increase in training, support, or compensation. The Apollo numbers give that feeling an actual dollar figure.

It is also a reasonable argument for tracking your own output and raise history more deliberately than you might have before, since the kind of wage compression this study describes is easy to miss year to year and only becomes obvious in hindsight.

The takeaway

The scariest AI jobs headlines are usually about mass layoffs that have not actually happened yet at the scale predicted. This study points at something less dramatic and, in a way, more immediately relevant to more people: a slow leak in pay growth that does not show up on a layoff tracker and does not make anyone update their resume. If you are in one of these roles, the practical move is not panic, it is paying closer attention to whether your compensation is actually reflecting the extra output AI is helping you produce, because on this evidence, a lot of employers are quietly pocketing that difference.