AINEWS
The Companies Spending the Most on AI Are Also Adding the Most Entry-Level Jobs
A joint paper from Ramp and Revelio Labs tracked 21,599 U.S. companies. Heavy AI spenders grew headcount 10.2% over 24 months, with entry level roles up 12%.
A joint paper from Ramp and Revelio Labs tracked 21,599 U.S. companies. Heavy AI spenders grew headcount 10.2% over 24 months, with entry level roles up 12%.
The companies with the heaviest AI bills are not laying people off. They are growing, and they are doing it at every rung of the ladder, including the entry-level one a new joint study says is most at risk from automation.
A paper published at the end of June by financial-operations platform Ramp and workforce-data firm Revelio Labs, "A New Look at AI's Impact on Jobs: Firm-Level AI Spending and Workforce Adjustment," tracked 21,599 U.S. companies by linking Ramp's corporate card and bill-pay data to Revelio's payroll records. Companies defined as "high-intensity adopters," the top third of per-employee AI spending in their first three months of paying for tools from vendors like OpenAI and Anthropic at roughly $33.67 per employee per month, grew headcount 10.2% over the 24 months that followed. The rest of the sample, where AI spending was $2.78 per employee per month, showed no statistically significant change in headcount, according to the paper and Ramp's accompanying letter.
Headcount for entry-level roles at high-intensity adopters grew 12% over the same 24 months, and the share of the workforce sitting at the entry level rose 1.15 percentage points relative to a control group of firms that had not yet started paying for AI tools. Low-intensity adopters did not gain entry-level share. They drifted slightly lower, the paper notes.
"The most striking piece is that entry-level workers at heavy AI adopters are growing in both absolute and relative terms," Ara Kharazian, who leads Ramp Economics Lab, told ZDNet. The pattern points at a mechanism the paper does not claim to have proven: companies that commit to AI may be selecting for and training a new kind of junior hire who can use the tools.
Heavy AI adopters were already larger, more engineering-heavy, and more than three times as likely to be venture-backed than the rest of the sample. Their median year-over-year headcount growth before they ever paid for an AI tool was 6.0%, against 1.6% for firms that never adopted, the paper notes. The 24-month post-adoption comparison corrects for the timing of adoption, not for the kind of company that adopts in the first place.
The 24-month window is short. The paper itself flags that it cannot resolve longer-horizon displacement effects. Ramp's Economics Letter adds a learning curve: headcount gains begin six to twelve months after adoption and compound from there. Whether they keep compounding past the two-year mark is the open question the data set cannot yet answer.
By the end of 2025, roughly a quarter of the sample met the threshold of three or more consecutive months of at least $100 a month in AI vendor purchases. The Information sector crossed 50% adoption. Finance and insurance, and professional and technical services, were close behind. Healthcare, construction, accommodation and food services, and arts and entertainment lagged, per the paper. Inside tech, California firms adopted more intensively than New York firms, and VC-backed companies of any sector were more likely to land in the high-intensity group than legacy tech firms.
Two things sit in the background: Ramp sells AI and finance automation tools, and Revelio Labs sells workforce data to employers, per the press release. Vendor-of-record interest is something to keep in mind.
The study lands inside a wider forecast range. Forrester projects roughly 6% of U.S. jobs replaced by AI by 2030, about 10.4 million roles. Boston Consulting Group estimates 10% to 15%. Anthropic CEO Dario Amodei warned in 2025 that up to half of entry-level white-collar jobs could disappear, and later walked that figure back. None of these forecasts operate on the same two-year post-adoption window the Ramp and Revelio paper covers. The two scales are different, and the gap between what adopting firms are doing now and what forecasters say could happen later is unresolved.
Revelio's own write-up of the joint study adds one more beat: small businesses are less likely to adopt, but when they do, they adopt more intensively per employee, and the marginal hiring effect is larger. Ramp has announced a partnership with Meta Small Business to push adoption further down the size distribution. That rollout will extend the dataset to a population the current paper underweights: small firms that have not yet had the runway to spend on AI tools.