The feared "AI jobpocalypse" keeps not arriving, and three very different pieces of research published this week help explain why the real story is stranger. AI is not making work disappear. It is quietly rearranging who does it, who trains for it, and above all who keeps the money when a smaller team produces more. The disruption is a question of distribution, not destruction, and the answer is turning out to be a choice rather than a fate.
Start with the clearest evidence. A study from King's College London and Stanford, drawing on 1.25 billion job advertisements and 158 million positions across 41 countries, found that when a firm adopts generative AI its total headcount actually rises, by about 3.3 percent. But the shape changes underneath: senior roles grow 6.7 percent while junior roles shrink 2.5 percent. The same pattern showed up from the United States to Brazil to Saudi Arabia. AI is absorbing the routine tasks that used to be the training ground for newcomers, and companies are responding by hiring more experienced people and fewer beginners. The bottom rung of the ladder is being quietly sawn off.
The mood inside offices matches that anxiety. A Visier survey found that nearly half of workers admit to exaggerating how much they use AI or how good they are with it, a phenomenon the firm calls "performative AI." Forty-five percent feel pressure to use tools they are not confident with, only 28 percent think their leaders fully understand how the technology is actually used, and 59 percent doubt their employer even has a plan. One laid-off software developer put it flatly: asked in interviews what excites her about AI, "I have to make up something, because frankly, nothing really excites me about AI." Beneath the enthusiasm is a trust gap, and the fear of being on the wrong side of the next cut.
And yet, for those who actually lean in, the tools deliver. A Gallup study found 63 percent of workers say AI helps them finish tasks faster and 46 percent say it raises the quality of their work, with regular users reporting far bigger gains than occasional ones. So the technology genuinely lifts output. The uncomfortable question, then, is the oldest one in economics: when a worker produces more, who gets the surplus?
A sweeping essay on India, co-authored with investor Saurabh Mukherjea, argues that the answer is political, not technological. Automation almost always displaces workers, it notes, but whether it also impoverishes them depends on who is in the room when the machines arrive. Germany gave its workers a legal voice over how technology is deployed and now runs one of the world's densest robot fleets with no net job loss. America dismantled that bargain, and since 1979 productivity has risen 94 percent while a typical worker's pay rose just 33 percent. India went further still: in its factories, labour's share of value added fell from 29 percent to 14 percent while profits climbed from 16 to 43, and two in five factory workers are now hired through contractors with no bargaining power at all.
The essay's warning is that white-collar India is now walking into the same trap one floor up, as entry-level hiring posts its first recorded annual decline and salaried work gives way to gig contracts. Stitch the three findings together and a single picture emerges. AI is not a jobs bomb; it is a lever that multiplies output and then hands someone the proceeds. The economists Daron Acemoglu and Pascual Restrepo call it a race between displacement and reinstatement. Who wins is not decided in a lab. It is decided by the rules we write around the machine.