After a year of headlines forecasting mass unemployment, a run of fresh data has arrived to say the opposite, and it is worth taking seriously. A Nomura study titled, pointedly, "Beyond the Headlines: AI Has Added Jobs in Asia, So Far," tallied 69 corporate announcements across the region since 2022 and found roughly 131,000 AI-related jobs created against about 61,000 lost. India, China and the Philippines all came out net positive. In the United States, a ZipRecruiter survey of more than 1,000 employers found 35 percent expect AI to grow their headcount over the next three to five years, against only 15 percent who expect cuts. The New York Federal Reserve's first "Street Level" report saw little sign of AI depressing employment even in the most exposed occupations. Sam Altman and Dario Amodei, who spent last year warning of white-collar carnage, have both softened their language.
So the ledger is genuinely positive. The trouble is that a net number can hide a wrenching redistribution underneath it, and this one hides two. The first is who the new jobs go to. Nomura describes a "K-shaped" labour market, where the fortunes of senior and junior workers diverge sharply. Employers want experienced staff who bring judgment, management and client relationships, and they are quietly closing the door on everyone else. In India, a survey of 651 IT firms found 55 percent had cut entry-level hiring, against just 14 percent trimming senior roles. In South Korea, employment among people aged 15 to 29 fell by 211,000, almost entirely in AI-exposed sectors, even as jobs for workers in their fifties rose by 209,000.
This is happening through subtraction, not spectacle. Rather than announce layoffs, companies simply stop hiring the young. Nomura calls it "silent firing," and its clearest illustration is India's IT sector, where the staffing firm Xpheno estimates net hiring collapsed from around 600,000 in the 2022 financial year to 140,000 in 2026, mostly through frozen campus recruitment. No headline, no protest, just a graduating class that finds the bottom rung has been sawn off. It is the least visible kind of disruption and, for a generation trying to start careers, potentially the most damaging, because the entry-level jobs being cut are exactly where people once learned to become the senior workers everyone now wants.
The second catch is what the new jobs actually are. Some of the employment the AI boom "creates" is the human labour that props the illusion up. Researchers studying data-labeling workers in China and Australia describe a global underclass drawing bounding boxes and annotating audio so that models can appear intelligent, often for as little as the equivalent of a few dollars a day, with no contracts, classified by platforms as "users" rather than employees, and nursing chronic eye strain from the hours. "We do the manual work so that they get the credit for the intelligence," one told the researchers. When a report says AI added jobs, some share of that total is this.
None of which makes the optimists wrong, exactly. AI really is proving to be more a transformer of work than a destroyer of it, and the apocalypse has not arrived. But "more jobs than it cuts" is a claim about a sum, and people do not live inside sums. They live at a particular age, in a particular role, on a particular rung. The most honest reading of this week's numbers is not that the fear was overblown, but that it was aimed at the wrong target. The danger was never a sudden collapse in the total. It is a slow, silent reshaping of who gets to work, and on what terms.