No one knows whether AI will trigger a white-collar jobpocalypse. The loudest warnings still come from people building and selling the technology, whose predictions often double as hype-mongering or cover for unrelated cost cutting with investor-friendly language. Think-tank and analyst forecasts are no less vertiginous.

The honest answer is that current data gaps leave us all guessing.

But we do know the script that’s already creeping up as displacement begins to trickle: “retraining,” “upskilling” and the hollow language used to make dislocation politically palatable. We cannot allow it to become the next comforting lie, or a way to sell the future while abandoning the people forced to live through it.

Policymakers and tech leaders shouldn’t wait for mass disruptions. They should also stop pretending retraining is much more than a slogan. America already ran this experiment after deindustrialization. Manufacturing losses were relatively concentrated, but the damage spread through families, communities and politics for generations. The payoff from the retraining policy during this era was weak at best.

The optimistic case is that AI will eventually create new jobs we can’t yet imagine, just as past tech revolutions did. But we won’t get there if policymakers ignore what Brookings Institution Senior Fellow Molly Kinder calls the “messy middle”: the period between today’s rocky AI adoption and the promised land of post-AGI abundance. Tech leaders love the utopian pitch, but the public may revolt long before it arrives. And retraining policy, Kinder notes, “has been one of the worst-performing categories of labor market intervention.”

The scale of potential labor exposure, meanwhile, dwarfs deindustrialization. Bloomberg Economics says that 27% of workers in advanced economies, more than 120 million people, are likely to be “meaningfully affected by AI.” That doesn’t mean all will be displaced, but it points to vast potential disruption.

Business leaders are already reaching for retraining. Nearly a quarter of CEOs in a recent survey said that more than half of their workforce will need to be “upskilled.” A banking chief went viral for his comments last month about AI replacing “lower-value human capital.” After the backlash, he apologized and invoked the familiar promise of reskilling.

Yet what does that mean? Too often, the answer is maddeningly vague. A recent analysis from human-resources consulting firm Randstad describes the future skill set as “AI fluency” plus “uniquely human capabilities” such as emotional intelligence, creativity, problem solving, critical thinking and ethical judgment. The first part is obvious; workers need to know how to use AI tools. The second part sounds like a list of desirable traits, not a transition plan for laid-off workers.

Inside the tech industry, the brazenly arrogant default answers are “universal basic income” or “learn a trade.” Blue-collar work may be less exposed to generative AI, by Anthropic PBC and OpenAI’s own estimates. But neither answer is complete. UBI requires faith that policymakers will build redistribution that doesn’t just reduce citizens to consumers. “Learn a trade” carries the same echoes of “learn to code” — a half-hearted retraining mantra turned weaponized meme that now looks grimly ironic as software engineering becomes one of the jobs most exposed to AI. It also assumes society will suddenly pay carpenters, butchers, plumbers or house cleaners far more than it does now.

People, meanwhile, aren’t just going to do nothing. In China, the OpenClaw mania we saw earlier this year was as much a story about scrambling to reskill as it was about agentic software adoption. In the US, anxiety is taking a more oppositional form, marked by communities protesting data centers. Both are versions of the same instinct as people see the future being built around them and try to regain agency.

Another hopeful case borrows from the University of Chicago economist Alex Imas’ influential essay asking “What will be scarce?” If AI makes cognitive tasks cheap, the idea goes, human relationships will become more valuable. Education, eldercare, social work and “relational sector” jobs requiring a human element should then command a premium.

Perhaps; but those are precisely the sectors that many societies undervalue. If policymakers want to turn them into the well-paid work of the future, they must invest now in more funding for care and education work, more apprenticeships, stronger career ladders and a broader effort to raise wages and dignity attached to these jobs. It also assumes that many people will be laid off first. Wages are unlikely to magically rise because out-of-work coders or marketers enter the field.

Any serious retraining policy must also give labor a real seat at the table. It’s lazy to assume workers are anti-technology. Union leaders often say that if technology makes jobs safer and easier, workers will support it. What they want is a say in how it’s used.

Ultimately, the transition period needs more than the false comfort of reskilling. It needs real-time data tracking on where jobs are being augmented or destroyed. From there, policymakers can focus on transition insurance, targeted public investments in care work and education, or a plan for sharing the productivity gains if AI abundance ever arrives.

Workers can adapt; people are endlessly resourceful. The risk is that “reskilling” becomes the excuse that makes mass unemployment politically palatable and, basically, the victim’s fault. If it’s being touted as the bridge to an AI future, it must lead somewhere. Otherwise, it’s just a signpost at the edge of a cliff.

____

This column reflects the personal views of the author and does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

Catherine Thorbecke is a Bloomberg Opinion columnist covering Asia tech. Previously she was a tech reporter at CNN and ABC News.

©2026 Bloomberg L.P. Visit bloomberg.com/opinion. Distributed by Tribune Content Agency, LLC.