For two years the loudest forecast about artificial intelligence and work was that white-collar jobs were about to vanish in a wave. This week the mood flipped. "The jobs apocalypse has been officially postponed," ran one widely shared summary of new analysis from The Economist, and the top-line numbers back the relief. AI has created roughly a million new US roles since mid-2023, against about 200,000 layoffs pinned to the technology, a ratio of five to one. AI-attributed cuts run near 16,000 a month, less than one percent of the 1.7 million job separations the economy churns through anyway. In August employers added 162,000 jobs and unemployment sat at 4.1 percent, lower than in nearly nine of every ten months over the past half-century. If AI were gutting the workforce, this is not what the data would look like.
It is a genuine correction to a genuinely overheated fear, and it deserves to be taken seriously. Morgan Stanley made much the same case this week, arguing that the college-educated, high-income, city-dwelling workers most exposed to AI are also the ones best placed to gain from it, through productivity, new roles and rising wealth. The New York Federal Reserve went further still, headlining its own survey "Businesses Are Using AI to Transform Work, Not Cut Jobs." Sixty percent of service firms now use AI, up from forty a year ago, and fewer than five percent say they have cut staff because of it.
But averages are exactly where the interesting damage hides, and the same week produced two findings that the cheerful headline glosses over. The first is the bottom rung. In New York, entry-level tech postings have halved since 2022, even as overall tech employment holds steady. "Entry-level postings have fallen off a cliff, and the first rung of the ladder is under pressure," said Jonathan Bowles of the Center for an Urban Future. The firms with the heaviest AI exposure cut the most junior roles, and those were the jobs that once paid newcomers around 89,000 dollars to start. Gartner warns the logic is short-sighted: it expects up to 30 percent of workers displaced by AI to be rehired by 2029, often at higher cost, once companies discover they have hollowed out the pipeline that produces their own experienced staff.
The second finding is subtler and may matter more. AI may be pressing on pay before it touches employment. A study from Apollo's chief economist found workers in highly AI-exposed occupations saw real wages grow 6.7 percentage points slower after 2023 than everyone else, with no measurable hit to headcount. The suggestion is that firms are pocketing the productivity gain through wage compression rather than layoffs. Economists rightly caution that the sample is thin and tech's post-pandemic hangover muddies the picture. Yet MIT's Daron Acemoglu expects the same: in a flexible labor market with a weak safety net, he argues, the impact will show up in wages before jobs.
The neatest illustration comes from David Autor's look at two once-similar clerical jobs. Accounting clerks lost a third of their positions to automation but saw wages climb 39 percent as the survivors grew more specialized. Inventory clerks went the other way: employment ballooned 175 percent while pay fell 13 percent. Same technological force, opposite fates. That is the real lesson buried under this week's good news. The apocalypse framing was always too crude, but so is the all-clear. AI is not deleting the workforce; it is quietly redrawing who gets in the door and what the work is worth once they do.