The Heaviest AI Adopters Are Hiring, and Entry-Level Is Growing Fastest

The story everyone expected: AI arrives, headcount falls, entry-level jobs disappear first. The data that actually showed up this summer says something close to the opposite.

A working paper from Ramp Economics Lab and Revelio Labs matches corporate spending records against workforce data for 21,559 American firms and finds that heavy AI adopters, companies spending about $34 per employee per month in their first three months of adoption, grew headcount 10.2 percent over the following two years. Low-intensity adopters saw no statistically significant change. The sharpest surprise sits at the bottom of the org chart: entry-level headcount at heavy adopters rose 12 percent, directly contradicting the assumption that automation clears out junior roles first.

The payoff runs on a lag

The gains take patience. Hiring effects don't show up until six to twelve months after adoption, then compound. A quarterly ROI review is exactly the wrong instrument for that curve: it reads a lag as a failure and recommends canceling right before the payoff arrives.

The macro backdrop makes the finding easier to trust. No general boom is doing the work here. Unemployment sat at 4.3 percent in May 2026, hiring is steady rather than hot, and consumer sentiment remains grim. Against a flat market, a double-digit headcount gap between heavy and light adopters is a real divergence. The authors flag their own caveats: the sample skews toward larger, technically sophisticated firms, and a 24-month window may miss longer-run effects. Even so, among the firms spending the most on AI, employment is rising, not falling.

The composition of the growth matters too. The gains cut across functions: engineering, sales, administration, and customer service all grew, though sector-level growth clustered in information businesses. The paper also notes that adopters were already larger, more technical, and faster-growing before adoption, which means heavy AI spending partly marks the kind of company that grows anyway. Even with that discount applied, the pattern is hard to square with the story of mass replacement, and it puts the burden of proof on the pessimists for the first time in three years.

The people hardest to evaluate on paper

For talent leaders, the second-order effect is the one worth planning for. If AI-intensive companies are expanding entry-level hiring, they're recruiting from precisely the population the traditional funnel evaluates worst. Early-career candidates have thin resumes by definition. A 22-year-old's one-page history of internships and coursework never predicted much, and now that the page itself is usually machine-polished, it predicts even less. Companies planning to grow at the entry level are planning, whether they know it or not, to make thousands of decisions the resume can't inform.

Some will solve this with pedigree, hiring from the same short list of schools and hoping the admissions office did the screening. That approach shrinks the pool, imports someone else's biases, and ignores decades of evidence that measured ability beats institutional brand as a predictor. The alternative is to evaluate the candidates directly: cognitive ability, learning agility, job-relevant skills, measured rather than inferred.

The Ramp findings undercut the AI jobs apocalypse, at least for now, and that's genuinely good news. But they also quietly raise the stakes on selection. The companies hiring fastest are hiring the people hardest to evaluate on paper. How well they choose will decide whether that 12 percent entry-level surge looks like an investment or a write-off two years from now.