When a job loses its easy tasks it can get harder and pay more; lose the hard tasks, and it can get easier and pay less — same automation, opposite outcome.

This explores whether automating part of a job leaves workers with harder or easier work. The corpus's short answer: it can go either way, and which tasks get automated decides it. Automating routine, low-expertise tasks leaves a job more specialized. Wages rise, but fewer people qualify for it. Automating the expert parts does the opposite. The job gets easier, pay falls, and less-skilled workers can enter Does automation raise or lower the skills that remaining work demands?. The odd result is that employment can move in the opposite direction from the number of tasks removed. A job that loses its expert tasks can end up employing more people, because more people can now do it.

A second line of argument explains why the remaining work often becomes more valuable, not just more demanding. In many jobs the tasks work like links in a chain: if one step is done badly, the whole result fails. Economists call this the O-ring model, after the part whose failure destroyed the Challenger shuttle. When some links are automated, the human links become the bottleneck, and worker time goes to them. Their quality, and their value, rises Does automating some tasks actually increase the value of remaining work?. Autor's classic example is bank tellers. Their employment grew while ATMs spread, because tellers shifted toward relationship and problem-solving work that the machines made more central Does automating some tasks make other human tasks more valuable?. Firm-level data adds a condition: when AI affects only a few tasks in a job, workers can move to the tasks that remain, and net job losses stay small Does concentrated AI exposure enable workers to adapt and reallocate?.

Sometimes the removed task doesn't disappear at all. It changes into something new. At a shipbroking firm, manual classification work wasn't eliminated when algorithms arrived. It was repurposed to produce the reference data used to audit the algorithm and then to adjust it. That loop became a skill the firm valued strategically Does automation eliminate manual work or repurpose it?. The remaining expertise here is a different kind, not just a higher level: knowing enough about the old task to judge the machine doing it.

That points to a tension the economics papers don't address. Routine tasks are often how people build the expertise the harder work needs. Anthropic's engineers reported large productivity gains, yet most could fully hand off only a small share of their work. They also worried that handing routine coding to Claude erodes the hands-on practice needed to catch its mistakes Does AI assistance erode the skills needed to oversee it?. Researchers studied by the CoCoA team sometimes kept tasks an agent could do adequately, because doing the work themselves builds their future ability Why do researchers refuse to delegate tasks to AI agents?. Nielsen disputes how this is usually measured. Taking the AI away to test what skills remain, he argues, creates a situation that never happens at work. The better question is what higher-level skills grow when the AI is always available Does removing AI tools actually measure real skill loss?.

One caution: removing a task doesn't automatically reshape a job. In a randomized trial across 66 firms, Copilot saved workers about two hours a week on email, yet the amount and type of their other work didn't change. Most work depends on coordinating with colleagues, and that hadn't been reorganized yet Does Copilot change what knowledge workers actually do?. So the expertise question has three parts: which tasks were removed, whether the freed time actually moves to harder work, and whether people still get the practice they need to do that work well.