Jobs, Rewritten: The Op‑Ed That Quietly Moved the AI Debate
Picture an electrician threading conduit through the exoskeleton of a new data center at dawn. The work is meticulous, noisy, undeniably physical. Down the road, a diner fills with people whose jobs look nothing like his—nurses finishing the night shift, an HR generalist editing onboarding materials, a real estate agent answering messages that arrived while she slept. Yesterday, an op‑ed landed in Washington that tries to connect all of them: a case, from Google’s chief economist Fabien Curto Millet and Cambridge economist Diane Coyle, that AI is not an executioner of occupations so much as a relentless editor of tasks—and that the real risk is not machines, but whether we teach people to work with them.
Their point is deliberately unromantic. In the long accounting of automation, it’s task lists, not job titles, that get rewritten. They remind us that among the 271 occupations recognized by the 1950 U.S. Census, automation truly erased only one—elevator operator—while the rest morphed. The radiologist parable, for years trotted out as proof that AI would empty out reading rooms, has turned out differently: imaging volumes rose, and what stuck were the judgment-heavy parts of the job—advising clinicians, speaking with patients, synthesizing the scatter of records into recommendations. The scan didn’t vanish. The role expanded around it.
That’s the thesis. The reason it matters is who is saying it, and when. This is a mainstream, policy-shaping signal: the chief economist of a leading AI firm, paired with a prominent academic, telling lawmakers and employers to stop treating labor market outcomes as weather and start treating them as design. If you’re a regular here, you already know the contours of this argument. But yesterday’s piece pulls the lens wide: beyond prompts and copilots, there’s the physical build‑out—the substations, chillers, racks—that powers AI. That build‑out is already creating demand for welders, electricians, HVAC technicians, and the craft workers who make digital capacity real. In other words, the so‑called “AI jobs” aren’t just model evaluators and prompt engineers; they’re also the tradespeople in steel‑toed boots, hired because inference takes electricity and electricity requires infrastructure.
Tasks Are the Unit of Change—and the Unit of Anxiety
Framing AI as a task reallocator sounds reassuring until you sit with the implications. Tasks are where wages and bargaining live. When the toolkit changes, the composition of a role changes, and with it the leverage inside firms. If AI takes the rote drafting off a lawyer’s desk, the work that remains looks more like negotiation, synthesis, and client stewardship. That can raise productivity and pay for the people able to step into that redesigned role. It can also strand those who were paid for their speed, not their counsel. The op‑ed is clear that this transition will be uneven. History’s scars—telecom switching, the wave of manufacturing automation in the late twentieth century—show what happens when communities are told to “reskill” without time, money, childcare, or credible pathways. The pain wasn’t inevitable; it was a policy choice.
Yesterday’s authors reject fatalism but not friction. They argue that net job creation is likely if we manage the bridge. That’s not boosterism; it’s a reminder that technology adoption spills into complementary demand. Data centers don’t just hire electricians; they pull in local logistics, security, maintenance, and the service economy. Hospitals that deploy AI in documentation don’t cut clinicians; they redirect them to patient‑facing work that had been squeezed. But complements don’t materialize on their own. Someone has to pay for the time it takes a mid‑career worker to learn a new workflow, and someone has to certify that the learning actually happened.
Design, Not Prediction
“The future is not a forecasting exercise — it’s a design challenge,” they write. That line lands because it names the bottlenecks we usually glide past. Training is not a press release; it is a schedule. Most of the people who will make up the 2030 workforce already have bosses, kids, commutes, and mortgages. Meeting them where they are means shifting from airy promises about online courses to concrete, employer‑led apprenticeships, paid on‑the‑job learning, and portable ways to document new skills. Without portability, firms underinvest because they fear poaching; without employer skin in the game, workers are asked to gamble their nights and savings on certificates that hiring managers barely recognize.
This is the part of the op‑ed that should make policymakers sit up. If you want AI to lift productivity and wages, you have to underwrite time. You have to reward employers for verifiable training—not the slide deck version, but the version that shows up in the tasks people actually do. You have to set standards so that a machinist in Ohio and a medical coder in Arizona can carry their new AI‑assisted competencies across companies and state lines without starting from zero. And you have to build the connective tissue: community colleges aligned with local employers; public procurement that prefers vendors who train and prove it; permitting and infrastructure plans that treat data centers as training anchors, not just tax assets.
What This Means for Readers Who Feel “Replaced”
If you’re waiting for the obituary of your job title, you may wait a long time. The replacement is more intimate. It arrives when a Tuesday task vanishes, a Wednesday task appears, and a Thursday task is suddenly measured differently. That is destabilizing even if the occupation’s nameplate stays bolted to the door. Yesterday’s argument offers clarity without comfort: most roles will persist, but the composition will tilt toward AI‑literate judgment and coordination. Those who can draft with a model, audit its outputs, and integrate its suggestions into the unwritten parts of work—client trust, team dynamics, ethical constraints—will see demand. Those who can’t will be boxed into the fragments that are easiest to price and outsource.
There’s also a geographic truth hiding in plain sight. AI infrastructure is landing where land, power, and permits are available, which is not always where displaced workers live. If the country treats training as a Zoom link and tells families to uproot without housing and school supports, we’ll replicate the worst chapters of prior transitions. If we treat training as local capacity—co‑designed with unions, employers, and educators, tied to real openings, delivered during paid hours—we can bend the curve. The same electrician outside that data center can become a foreman on AI‑driven energy projects; the same nurse buried in paperwork can spend more minutes at the bedside because documentation runs faster. Those outcomes are designed, not decreed.
The Near Horizon
Over the next year, expect a premium on workers who treat AI as a collaborator rather than a novelty. Expect hospitals, law firms, and accounting shops to quietly rewrite job descriptions, swapping “proficient in X software” for demonstrations of how someone supervises AI outputs in messy, real contexts. Expect a lot of branding around “skills,” and a smaller set of institutions that actually verify them. Watch whether states stand up registries for portable credentials, whether employer coalitions report completion and placement rates, and whether big AI infrastructure projects include funded apprenticeship slots instead of photo‑ops.
The op‑ed’s bet is that labor demand remains strong if the bridge holds. That rings true. But bridges fail for boring reasons: nobody budgeted for paid learning time; a credential meant one thing in one region and nothing in another; a cohort was accepted but childcare never materialized; a firm used the tool to squeeze throughput without redesigning the job. The authors have given Washington a new focal point. The rest of us will know it’s real when a mid‑career worker can show a verifiable “skills transcript,” cross the street to a better role, and be welcomed in without having to start from scratch.
Jobs aren’t disappearing. They’re being rewritten. The only question is whether we write them with the people who do them—and whether we pay for the edits.
