There's a specific kind of anxiety making its way through open-plan offices and home workspaces alike. It doesn't announce itself. It hums beneath the status meetings, the Slack notifications, and the chirpy all-hands emails about "exciting changes ahead." It's the feeling of watching the ground shift under your feet while everyone around you pretends the floor is perfectly still.
White-collar workers are scared. Not of a pink slip landing on their desk — that fear has a shape, a timeline, a thing to fight. This is murkier. It's the fear of slow erasure.
Meet the new precariat. They have degrees. They have mortgages. They have titles that used to mean something. And increasingly, they're spending their workdays supervising software that's learning to do what they spent years learning to do. A data scientist in New York watches AI agents absorb her codebase. A copywriter in Brooklyn watches language models churn out in seconds what used to take her hours. An insurance tech worker in North Carolina builds automation flows and can't shake the feeling she's constructing her own replacement, brick by brick.
This isn't the automation story we were told to expect. The old script had robots coming for factory workers, truck drivers, warehouse crews — the physical, the repetitive, the easily mechanized. The knowledge worker was supposed to be safe. Creativity, judgment, nuance: these were the moats. Now those moats are being quietly drained.
The cruelest part isn't the uncertainty. It's the non-answer. Previous generations facing economic disruption at least had a ladder to climb. Get educated. Learn a trade. Move to where the jobs are. The advice was often inadequate, but it existed. Today, no one credible can tell you which skills to build, which industries to plant your flag in, or what the job market looks like in five years. Tech leaders oscillate between utopian and apocalyptic with the confidence of people who will be fine either way. "Jobs will transform," they say, with a vague wave of the hand. Into what, exactly? By when? Silence.
That ambiguity is doing real damage — not just to careers, but to people. Researchers have a name for chronic job insecurity: it's a diffuse psychological condition that erodes your sense of control and, with it, your ability to act. It manifests as anxiety, burnout, and depression. A recently coined term — "AI replacement dysfunction" — describes the particular paralysis of workers who can feel displacement coming but can't calculate how close it is.
The economy is doing this on purpose. That sounds conspiratorial, but it isn't. It's just how the incentives are structured. Corporations aren't deploying AI to enrich their workforce — they're deploying it to cut costs, pad margins, and reward shareholders. The workers being replaced, deskilled, or surveilled aren't a side effect. They're the point.
Middle managers are pressured to report AI-driven successes that haven't materialized. Algorithmic monitoring tracks keystrokes, flags "at-risk" employees, and predicts performance in ways workers never consented to and often don't know about. Emails arrive from HR assuring employees that AI is a tool, not a threat — written, employees note, in language that sounds unmistakably machine-generated. The irony is thick enough to choke on.
What's getting lost in all of this isn't just productivity or efficiency. It's something harder to quantify: the experience of doing skilled work well. The flow state. The satisfaction of wrestling a complex problem to the ground. The craft. Companies are stripping out the very elements of work that make workers good at their jobs — and in doing so, they're hollowing out the institutional knowledge and human judgment that made those companies worth anything in the first place. Bots can generate. They can't care.
This isn't new, but the speed is. A century ago, Taylorism broke skilled tradespeople into interchangeable parts. Thinking was reserved for management; labor was for following instructions. Autonomy disappeared. Morale followed. What's happening now is the same logic migrating up the org chart, colonizing roles that were supposed to be immune. The office is becoming a kind of factory floor for cognitive labor, with all the alienation that implies.
Globalization shipped jobs overseas. Financialization extracted value from the people who created it. Technology reshaped labor markets with each wave — the assembly line, the computer, the internet. AI is the latest disruption, but its speed is qualitatively different, and the existential edge to this one — what does it mean to be human if machines can think? — cuts deeper than previous shocks.
The way out isn't individual. It can't be. You can't hustle your way out of a structural crisis. You can't learn enough new skills fast enough to outrun a technology that learns faster. The workers who recognize this earliest are the ones quietly moving toward collective solutions.
Union approval in the U.S. recently hit its highest level in six decades. Entertainment workers have struck over AI. Telecom unions have negotiated AI principles. Across industries, workers have won advance notice requirements, human oversight mandates, and explicit protections against automatic replacement. None of it is enough. All of it points toward what enough might look like.
There's also something historically significant in who's feeling squeezed right now. White-collar workers have long held themselves apart from blue-collar organizing — different class, different risks, different culture. But the AI squeeze doesn't respect those distinctions. The "middle precariat," facing the same destabilization that's defined working-class life for decades, might finally have the motive to build something across those old divides. The New Deal coalition — messy, imperfect, but real — emerged from a moment not unlike this one: shared economic shock breaking down the walls between groups who had previously kept their distance.
What's actually needed isn't complicated, even if it's politically difficult. Workers need real retraining — not webinars, not "upskilling" brochures, but funded, substantial preparation for a changed economy. They need safety nets strong enough to absorb the disruption that's already happening. They need a seat at the table where AI policy gets made — not as stakeholders to be consulted and ignored, but as participants with actual power.
The government has to set rules and enforce them. When the federal government fails to act, states must move — and many are already trying. None of this happens on its own. None of it happens through optimism alone.
The people building the AI economy love to talk about the future. They are somewhat less enthusiastic about who gets to live well in it. That's a question workers, communities, and governments have to answer themselves — before the answer gets made for them.