Does cutting easy tasks from a job make it harder (and higher-paid), while cutting hard tasks makes it easier (and lower-paid)?

This explores whether automating part of a job helps or hurts workers depending on what's left over: harder, more expert work, or easier, more routine work. The corpus says yes, and the direction matters a lot. Does automation raise or lower the skills that remaining work demands? sets out the core result. When automation removes the easy, inexpert tasks, the job that remains needs more expertise. Wages rise, but fewer people qualify, so employment shrinks. When automation removes the expert tasks, the remaining job gets easier. Wages fall, but more people can do it, so employment can grow. The surprise is that employment moves opposite to what you'd expect from counting tasks. Losing tasks can open a job up to more people, and keeping the hard ones can shut people out.

A second idea complicates the simple 'tasks removed means jobs lost' picture. Does automating some tasks actually increase the value of remaining work? points out that many jobs work like a chain, where overall quality depends on every link holding. In jobs like that, automating some links frees workers to concentrate on the links that remain, so each remaining task becomes more valuable. Standard 'AI exposure' scores treat each task as having a fixed value, which means they likely overstate how many workers will be displaced. Read together with the first note, the question becomes less about how many tasks are removed and more about how the leftover tasks fit together and how much skill they need.

Skill can also shift at the level of individual workers. In Does AI help lower-skilled workers catch up to top performers?, GPT-4 helped weaker consultants most and narrowed the gap with top performers. In effect, AI did some of the expert work for them, which is the 'lower the bar' scenario from the first note. The same note warns that when people choose for themselves whether to use AI, the effect can reverse. Will AI increase inequality less than past automation? reaches a similar conclusion across the whole economy: AI's wage effects spread more evenly across groups than earlier waves of automation did, yet income from owning capital still pulls ahead of income from wages. How the task mix changes decides who wins, not only whether anyone does.

There is also a cost that wage models leave out: what the remaining work feels like. Does AI collaboration drain motivation when workers return to solo tasks? found that once AI had absorbed the engaging parts of a task, people who went back to working alone felt more in control but less motivated and more bored. The leftover work can be 'easier' and still be worse to do. Does removing AI tools actually measure real skill loss? adds that if AI stays in the workflow for good, the useful question is not whether people lose skills when the tool is taken away. It is which higher-level skills grow around the tool.

The corpus gives a clear conceptual answer but little direct evidence about which way real occupations are actually moving. Telling whether a given job is shifting toward 'harder residue' or 'easier residue' is still largely an open question.