The fear around generative artificial intelligence has been easy to summarize: software gets smarter, white-collar jobs disappear. A new working paper from Ramp and Revelio Labs complicates that story. In “A New Look at AI’s Impact on Jobs,” Ara Kharazian, Lisa Simon, and Ryan Stevens matched Ramp corporate card and bill-pay data with Revelio Labs workforce records for 21,559 U.S. firms. Their headline finding is blunt: firms making the largest AI investments grew employment by roughly 10% after adoption, while low-intensity adopters showed no statistically significant change.
What the spending data shows
The researchers sorted firms by AI spend per employee in the first three months after adoption. High-intensity adopters were the top third of spenders, at roughly $30 per employee per month at the outset and rising thereafter. Those firms grew headcount by about 10.2% over the two years after adoption, while low-intensity adopters saw no statistically significant change. The gap is the story.
The pattern at the entry level stands out. Entry-level headcount grew about 12% at high-intensity adopters, and entry-level workers’ share of the workforce rose about 1.15 percentage points relative to the control group. That cuts against the fear that AI will eliminate the bottom rung of the career ladder first. Gains were broad across engineering, sales, administration, and customer service, though the paper also says sector-level gains were concentrated in information firms.
There is a wrinkle that helps explain why some AI investments look underwhelming at first. Hiring gains did not appear until 6 to 12 months after adoption and then compounded over time. Buy the seats in January, and the org chart may not bend until the fall. A quarterly measurement window can be too short to capture the learning curve.
The macro backdrop
The study lands in a labor market that is steady rather than booming. The unemployment rate was 4.3% in May 2026, and nonfarm payrolls reached about 159.0 million. Job openings rose to 7.594 million in May, while initial jobless claims were 215,000 for the week ending June 20. The real-time Sahm Rule recession indicator was 0.10 in May, well below the 0.50 recession threshold, and average hourly earnings reached $37.53, up 3.4% from a year earlier.
The tension shows up in household mood rather than only in the spreadsheet. The University of Michigan Consumer Sentiment Index fell to 44.8 in May 2026, down from 61.7 in July 2025, before rebounding to 49.5 in June. Workers are getting hired and paid more while many households still feel worse about the economy. AI fear may be part of that mood, but inflation and the cost of living remain central drivers.
What it means for actual people
Kharazian draws three practical takeaways from the data:
- Young workers and recent graduates choosing between otherwise similar employers should pay attention to which one uses AI seriously, because entry-level hiring grew faster at high-intensity adopters.
- Engineers worried about automation should note that those adopters also expanded engineering headcount.
- Business owners who have not yet seen gains should remember that the paper found both a spending threshold and a learning curve.
The caveats matter. This is an early working paper, and the findings describe firms in aggregate, not any individual job. AI adopters were already larger, more technical, faster-growing, and more likely to be venture-backed. Adoption also clusters through networks: Ramp says funding source predicted adoption better than sector, California-based tech firms over-indexed, and small businesses adopted less often but more intensively when they did. The signal to watch is whether entry-level hiring at heavy AI adopters keeps outpacing the rest of the labor market as cheaper tools spread across the economy.
The Next Test for the AI Jobs Story
The cleanest reading is not that AI is saving every white-collar job, or that layoffs blamed on AI are imaginary. It is that serious AI adoption may help some firms expand faster after a learning curve, especially in technical and information-heavy sectors. That is a narrower claim than the hype version, but it is also more useful. The next question is whether the hiring gap survives once AI tools become ordinary software rather than a marker of aggressive, well-funded firms.
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks