the-impact-of-automation.svg" alt="The Impact of Automation" width="1200" height="800" loading="lazy" decoding="async">Every wave of automation has produced two predictable responses: a wave of optimism about freeing humans from drudgery, and a wave of pessimism about the jobs that will vanish. Both responses are usually right. Both are usually wrong about which specific jobs they're describing.
The historical pattern is that automation eats tasks, not jobs. A job is a bundle of tasks; automation peels them off the bundle one at a time. The radiologist who was supposed to be replaced by image-recognition models in 2018 is still employed in 2026 — but spends more of her day on ambiguous cases and less on routine reads. Net, she's busier than ever. The juniors who would have done the routine reads are the ones absorbing the displacement.
The pattern this time looks similar, but with a wrinkle: the tasks being automated now are cognitive ones, and they're being automated faster than the labor market can re-skill around them. The transition cost falls disproportionately on people in the early- and mid-career bands.
The policy answer is the boring one — better safety nets, real re-skilling infrastructure, less ideological hand-wringing — but it's been the right answer through every prior wave too. Automation is not new. Pretending it is is what gets us in trouble.
The Historical Pattern
Automation isn't a new phenomenon. The Luddites of the early 1800s were skilled textile workers smashing power looms because the looms were eliminating their trade. The displaced agricultural workforce of the early twentieth century is one of the most documented economic transitions in history. The middle class created in the postwar period was, in significant part, the workforce that was displaced from agriculture and reabsorbed into manufacturing. Each wave produced exactly the dislocations and adaptations that the current wave is producing, on roughly similar timescales.
What the historical pattern shows clearly is that the labor market does eventually absorb the displacement — but slowly, and at substantial human cost to the workers caught in the transition. The grandchildren of displaced textile workers became factory workers. The grandchildren of displaced farmers became urban professionals. The transitions are not zero-sum across centuries, but they are deeply uneven within them. The question for our current wave is whether the absorption timeline is compatible with the political stability of the societies absorbing it.
What's Different This Time
The standard automation skeptic position is that "this time is different" is the most overused phrase in economic history, and every wave of automation has produced the same fears, all of which have been disproven by eventual labor-market absorption. There's truth to that. But there's also a specific way in which this wave is structurally different from previous ones, and pretending otherwise is its own form of intellectual laziness.
Previous waves automated physical labor and routine cognitive labor. The displaced workers were absorbed by sectors that required the human skills the machines couldn't replicate. The current wave is automating non-routine cognitive labor — writing, summarizing, coding, designing, planning — which is the category of work that absorbed the displaced workers of every previous wave. The sectoral exit ramps that existed in 1900 and 1960 may not exist in the same form in 2030. That's not a guaranteed catastrophe, but it's a reasonable concern that deserves engagement rather than dismissal.
Where Policy Could Help
Most thoughtful policy responses converge on a few patterns. Universal basic income is the most discussed, but the evidence on it is mixed and the political appetite is limited. More incremental approaches include expanded earned-income tax credits, sectoral re-skilling programs with real funding behind them, and portable benefits that detach health care and retirement from employer-based delivery.
The least defensible policy response is the one most governments seem to default to: do nothing, hope the market sorts it out, and react to political backlash with rhetorical concessions rather than structural ones. Historically, this approach has resulted in the political instability that the same governments later complain about. Automation displacement, handled well, is a normal economic transition. Handled badly, it's the catalyst for the populist movements that define generations.
Individual Strategy
At the individual level, the most durable advice is some version of: develop skills that compound, not skills that commodify. Writing, judgment, taste, persuasion, and the ability to work across multiple domains are all examples of skills that remain valuable even as the cheap baseline of any given task gets automated. The specific syntax you learned in 2015 is mostly obsolete; the ability to think clearly about what to build remains valuable.
The harder advice is that almost everyone reading this is going to have to retrain at least once in their working life, and possibly multiple times. The career arc of the postwar middle class — train for one job, do that job for forty years, retire — is no longer the modal career arc. It hasn't been for a couple of decades. Pretending otherwise produces worse outcomes than accepting it does. The people I know who handle the transitions best are the ones who built the habit of learning new things long before they were forced to.
What I Keep Coming Back To
What worries me most about the current automation wave isn't the loss of jobs in any specific industry. It's the speed of the transition. Previous waves played out over decades and gave the labor market time to absorb. The current wave is playing out over years, and the absorption mechanisms — re-skilling, sectoral migration, wage adjustment — operate on the older timeline. The mismatch is where the political instability comes from.
The way out, to whatever extent there is one, runs through institutions that aren't currently set up to move fast either. Community colleges. Workforce development boards. Unemployment insurance systems. State licensing regimes that still require classroom hours measured in semesters. Updating all of these to the actual pace of labor-market change is the unsexy, unfundable, generationally important work that almost nobody is doing. The people who do that work in the next decade will quietly determine how this generation experiences the transition.