Caretakers to the Machine

Some fifteen years ago, I offered a couple of predictions to a roomful of my graduate students. Then, because I was younger and believed the internet was forever, I went home and wrote about them on a blog that no longer exists. So much for the permanent record.

My first prediction was that the skilled labor market of the future would consist primarily of caretakers to intelligent machines—humans who did not perform the tasks but supervised them, made corrections, and dealt with the liability when the machine failed.

My second was about music. I said that the moment AI-generated tracks (which didn’t exist yet) became “good enough” to make up a meaningful share of consumption on a streaming service, the service would promote them to reduce its licensing costs, because it owes nothing to a rights holder that does not exist.

These scenarios have more or less arrived, faster than I anticipated, and I caught news about each in the same week.

First, the professional services giant PricewaterhouseCoopers, PwC, just analyzed more than one billion job postings for its 2026 Global AI Jobs Barometer. They found something concerning, especially for younger job seekers: employers now want entry-level workers with skills that used to take years to develop. PwC calls it “seniorization.” Those first-rung jobs, in occupations highly exposed to AI, are now seven times more likely to demand skills that appear much later in a career, like strategic decision-making, stakeholder management, leadership, and judgment.

Take a law firm. It would once have kept a stable of young associates doing the grunt work of reviewing case law, pulling filings, and crafting summaries. So much of that can now be done by machines that mostly you need somebody to double-check the work. Then you need more experienced staff to manage clients who now expect work done faster and at higher volume. What you no longer need are fledgling attorneys and paralegals of the traditional variety.

This is in line with my forecast. AI sands away the routine cognitive tasks— literature review, data cleanup, executive reports—all the grunt work nobody loved but that was part of the slow path by which a 23-year-old graduate could become a 35-year-old decision-maker. Now that capacity, plus the aptitude to manage new information systems, is in demand on day one. The end result could be manufactured scarcity, because the firms are not training up the workers who develop these competencies through organic, experiential processes.

My second prediction showed up in the same news cycle. The streaming service Tidal announced it will demonetize any track it identifies as wholly AI-generated and tag the rest, which sounds like a good thing. But Deezer, another streaming service, has already reported that AI-generated songs make up half of its daily uploads. That leaves a platform sitting on a swelling catalog it pays no royalties for, music surely “good enough” for playlist filler—the grind playlist, the workout playlist, the chill playlist, etc.

Once again, the surplus goes somewhere other than to the people whose labor built the economy in the first place. None of this is new. It is a very old pattern, and to find its beginning you have to go further than the Wayback Machine can take you.

The medieval plow increased agricultural output across Europe. But the value it produced went to build cathedrals, not to feed the peasants who worked the fields. The spinning jenny automated textile production but delivered longer hours at lower wages to the very people it was meant to liberate.

In modern times, the internet created more wealth than any technology in history and concentrated it among a handful of platform operators. Many workers—including trades and creative professionals—were shunted to gig roles and delivery routes.

The real question is whether these disruptions actually created value, or merely concentrated it as easy pickings for the ownership class.

The dynamic is not new, but the speed sure is. The plow took centuries to reorganize who eats and how much. It was only decades with the spinning jenny’s labor reset. Now there’s a rapid, wholesale change in who even gets to enter the employment market, across a broad band of occupations at once, with no plan for apprenticeship and no policies in place to soften the blow.

We are all being asked to welcome this mighty rearrangement as opportunity rather than the latest turn in the oldest scam, the one in which the machine performs exactly as designed and the benefit, as it always has, flows in a single direction.