A new study tracking AI adoption and hiring across more than 21,000 U.S. firms suggests the relationship between artificial intelligence spending and employment is considerably more conditional than either side of the debate tends to admit. Through May 2026, companies had announced close to 90,000 job cuts tied to AI, and projections from some quarters put the share of U.S. jobs at risk over the next five years at up to 15%. The picture from firm-level data, however, is harder to summarise than either a jobs apocalypse or a tech-fuelled hiring boom.
What AI Adoption and Hiring Data Actually Show
The paper, titled ‘A New Look at AI’s Impact on Jobs: Firm-Level AI Spending and Workforce Adjustment,’ was authored by Ara Kharazian and Ryan Stevens of Ramp and Lisa Simon of Revelio Labs, and published on 30 June 2026. It covers 21,559 U.S. firms, drawing on Ramp card and bill-pay records linked to Revelio Labs workforce data spanning 2021 through early 2026.
The core finding is this: companies the paper classifies as high-intensity adopters, defined as those spending an average of $30 per employee per month on AI tools in their first three months, saw total headcount grow by 10.2% over the subsequent 24 months. That growth spread across engineering, sales, administration, customer service, finance, marketing, and scientific roles.
The entry-level picture is equally striking to those who expected the opposite. The Goldman Sachs analysis, cited in the paper’s broader context, found AI has erased roughly 16,000 net jobs per month over the past year, with Gen Z and entry-level workers bearing the heaviest burden. Yet among high-intensity AI adopters in the Ramp-Revelio study, entry-level headcount rose by 12%. According to Revelio Labs coverage of the findings, the share of entry-level workers in high-intensity adopters’ workforces increased by 1.15 percentage points compared with organisations that had not yet adopted AI, while companies making smaller investments recorded a modest decline in that proportion.
Low-intensity adopters fared differently: the paper found no statistically significant change in headcount at all. Not slower growth, but no detectable movement. That gap between committed spenders and tentative experimenters runs through every finding in the study.
The Limits of the Optimistic Reading
The authors are direct about what the data cannot claim. ‘This paper does not show that AI universally creates jobs,’ they write, ‘but it does counter claims that AI will lead to broad job losses.’ The sample skews toward tech-forward, knowledge-work firms, many of which may have been on a growth trajectory regardless of their AI spending. Disentangling AI’s contribution from the underlying momentum of a venture-backed software company expanding into new markets is, by the authors’ own account, genuinely difficult.
The mechanism the paper proposes is specific to that kind of firm. ‘For software and technology firms, AI can make core output cheaper or faster to produce: writing code, debugging, building internal tools, producing technical documentation, and supporting product development,’ the report reads. ‘Lower production costs in these workflows can raise the return to expanding the whole firm, not just the engineering team.’ In other words, AI is functioning as a growth accelerant in these companies, not a headcount trimmer.
The broader Goldman Sachs picture adds useful texture. Economists Sarah Dong and Joseph Briggs, cited by Fortune, estimate that data centre construction alone has added 212,000 jobs since 2022 and continues generating around 9,000 new positions per month. Separately, Goldman calculates that AI augmentation has added roughly 9,000 jobs per month in occupations with high augmentation potential. Set against the net displacement figure of 16,000 jobs per month, and the finding from Goldman Sachs research cited by Yahoo Finance that AI has nudged the U.S. unemployment rate up by 0.1 percentage points, the net effect so far looks less like a cliff edge and more like a slow structural redistribution.
The inequality the Ramp-Revelio study flags may prove the more consequential story. Firms that bought subscriptions and ran pilots without sustained investment saw none of the hiring gains. The capital, technical staff, founder networks, and management bandwidth required to convert AI spending into genuine business expansion are not evenly distributed. As the paper’s authors put it: ‘Firms without those channels may fall behind.’
The Ramp press release accompanying the study frames the findings as an argument for deeper investment rather than cautious experimentation. Whether companies outside the tech-forward cohort can replicate those results is the question the next round of data will have to answer.
