The debate over AI and jobs has been frustratingly binary. Either AI destroys jobs or it creates them. The real picture is messier, and it splits along a line most companies do not want to look at closely: how seriously they actually use the tools.
Goldman Sachs economists, working from real payroll records, found that AI is currently erasing about 16,000 net U.S. jobs per month. That is 25,000 jobs removed by automation, with roughly 9,000 new ones added back through roles that work alongside AI. Entry-level and younger workers are taking the largest share of those losses.
A Harvard study covering 65 million workers across more than 280,000 U.S. firms confirmed the same pattern from a different angle. At companies that actively adopted AI, junior hiring fell sharply while employment at senior levels stayed essentially unchanged. Separately, entry-level job postings across the U.S. are down 35% since early 2023, with drops as steep as 67% in software and data roles.
Then comes the Ramp and Revelio Labs report, and it complicates all of that. Among companies spending roughly $30 per employee per month on AI from the start, total headcount grew by more than 10%. Entry-level headcount at those same firms rose by 12%. The gains showed up across sales, finance, customer service, marketing, and administration, not just engineering.
The reason this happens in some companies and not others comes down to economics. For technology and software firms, AI makes their core product cheaper and faster to build. That does not just save money; it raises the return on growing the whole business. Lower costs can mean lower prices, which can mean more customers, which can mean more hiring across every team.
But that logic only works if you have the right conditions in place. Companies that run pilots and pay for subscriptions without committing further see none of those hiring gains. The report is explicit: sustained, deep investment is what separates firms that expand from firms that just experiment.
A February 2026 survey of nearly 6,000 executives across the U.S., U.K., Germany, and Australia found that about 70% of firms say they "actively use AI," but executives spend an average of just 1.5 hours per week using it. About 90% of those firms reported no measurable impact on employment or productivity over the prior three years.
That gap between claiming adoption and actually embedding AI into operations is the core problem. And it is not evenly distributed. In the EU, large enterprises adopt AI at about three times the rate of small ones. The firms best positioned to close that gap are already-well-resourced: they have technical staff to implement properly, management capacity to redesign workflows, and capital to sustain the investment past the trial phase.
For operators outside the tech sector, the picture this paints is straightforward. One-off subscriptions and pilot programs are unlikely to produce the staffing or productivity gains the headlines promise. Industries like construction, hospitality, and agriculture sit at the lowest adoption rates globally, which means the gap between them and their more AI-intensive competitors compounds every year.
The data also raises a practical talent pipeline concern that goes beyond any single company. A global survey by Oliver Wyman found more than 40% of CEOs plan to cut junior positions in the next two years, reversing what had been a clear bias toward growing entry-level ranks just a year prior. Companies cutting the bottom of their workforce today are also cutting the people who would have become their mid-level managers five years from now.
The firms that treat AI as a genuine operational change, not a cost-cutting shortcut or a subscription to try out, are the ones seeing headcount rise. That is the narrow corridor where the optimistic case for AI and employment actually lives. For everyone else, the more cautious numbers are probably closer to the truth.