A couple of years ago the confident line was that half of office jobs would be gone by now. That window has closed, and the jobs are mostly still here. It’s worth sitting with why the scariest version didn’t happen, without pretending the past few years have been painless.
The prediction that didn’t land
The clearest fact to start with is a boring one. The US has added roughly 3 million white-collar jobs since late 2022, according to a January 2026 Economist analysis. That’s a workforce that kept expanding right through the period when the models got good enough to write emails, draft code, and summarize documents faster than most people can read them.
Google’s James Manyika put the mismatch plainly. “Some of those predictions were made two years ago,” he said, “that in two years, 50% of jobs would be wiped out. Well, two years is up.” His own read is that jobs are harder to automate than a lot of Silicon Valley assumes. His numbers help explain why: about half of work tasks are automatable, but only around 10% of occupations could be automated all the way through.
That gap between tasks and jobs is the whole story. Almost every job is a bundle of things, and a model can be great at three of them and useless at the other seven. Automate the three and you haven’t removed the job. You’ve changed what the person spends their day doing.
Wider, not gone
The Economist framed the likely path as expansion rather than erasure. “Rather than wipe out office jobs,” it argued, “artificial intelligence will expand their scope and raise their value.” I find that more convincing than the extinction case, and the mechanism is simple enough to picture.
When a chunk of your work gets cheaper to produce, you don’t usually go home early. You take on more. The marketer who used to spend two days on a first draft now spends two hours, then fills the rest of the week with campaigns, analysis, and calls that never fit before. The analyst who used to hand-build one model builds five and spends the freed time arguing about which one to trust. The job absorbs the slack. Its edges push outward.
This is also why the “cyborg” image keeps showing up in this debate. The office that emerges from all this looks less like an empty floor and more like people working alongside a fast, tireless, occasionally wrong assistant. The human stays in the loop because someone has to decide what to keep. That role doesn’t shrink the job. It’s arguably the most valuable part of it.
The part I won’t wave away
None of this means the ride has been smooth, and I don’t want to sell a clean story here. The aggregate held up, but averages are good at hiding who paid the cost. The people who absorbed most of the disruption are the ones with the least room to absorb it: workers just starting out.
Young workers in AI-exposed fields have been hired noticeably less over the past few years, even as the overall white-collar count rose. The routine tasks that used to fill a junior’s first year are exactly the ones a model handles well, so the on-ramp got narrower right when the highway stayed open. You can believe both things at once. The sky isn’t falling on office work as a whole, and the entry-level door has gotten harder to walk through.
So the optimism has to be the measured kind. “White-collar work survives” is true and also cold comfort if you’re 23 and sending your fortieth application into a void. The macro number and the personal experience are both real, and they point in different directions.
What measured optimism looks like in practice
If the jobs are mostly staying and the roles are getting wider, the useful question isn’t whether to brace for extinction. It’s how to be the person a wider role wants.
Wider roles reward range. When the routine parts get automated away, what’s left leans toward judgment, coordination, and knowing which output to trust. Those are the parts that don’t compress well, and they’re the parts worth building on purpose rather than hoping to pick up by osmosis.
They also reward people who can work with the machine instead of around it. The cyborg framing cuts both ways: the workers who thrive are the ones comfortable delegating the grunt work and then catching what the model gets wrong. That’s a learnable habit, and it’s becoming a hiring signal in its own right.
And a lot of this gets tested in the interview, because it’s hard to screen for on a resume. A hiring manager who wants someone with judgment will ask you to walk through a messy call you made and then push on the parts you skip over. Being able to hold that conversation, to reason through a problem under a little pressure without freezing, is a skill like any other, and it responds to practice against something that pushes back rather than nods along.
The honest read
The mass-unemployment forecast for office work was too confident, and the deadline came and went without it. The more likely future is roles that grow to cover more, not roles that disappear. That’s the good news, and it’s real. The catch is that the transition has landed hardest on the youngest workers, and no amount of macro cheer fixes that for the person living it. Both can be true. The safe bet is to stop waiting for the collapse and start building the parts of the job a machine can’t quietly take.