Work, Wages, and Who Absorbs the Adjustment

The AI transition reaches households through tasks, not through a single “exposure” score. This chapter examines how automation and assistance change the expertise content of work, why wages and employment can move in opposite directions, and how early labor indicators should be interpreted without turning them into a definitive forecast. The practical judgment is to track the task being changed and the timing of adjustment, then test whether support helps people move before the target work itself shrinks.

The question is not simply how many jobs AI might remove. An occupation is a bundle of tasks, and a technology can automate some of them while complementing or intensifying others. That internal change affects who can enter, what expertise is rewarded and how much labor the occupation needs. It also determines whether a later aggregate productivity gain arrives with higher incomes, weaker bargaining power or a prolonged period of insecurity.

The task, not the job title, is the mechanism

The IMF's task-based account begins with a distinction between the kind of work being automated and the expertise it requires. If AI automates a high-expertise task, the barrier to entering the occupation can fall. More workers may be able to perform the remaining work, increasing employment or widening the supply of labor, while wages stagnate or decline. If AI automates a low-expertise task, the work left behind can become more expert-intensive. The remaining workers may earn more, but the occupation can employ fewer people.

These outcomes are not contradictory. They follow from the direction in which the task mix moves. Removing a difficult task can make an occupation easier to enter; removing an easy task can leave a more demanding occupation. A headline that reports “AI exposure” without identifying this direction cannot tell us whether the main effect is substitution, complementarity, lower entry barriers or greater demand for expertise.

Consider the supply response. When an automated task was previously a barrier to entry, removing it can allow more people to do the occupation. The resulting increase in labor supply may raise total employment even if each worker earns less. When an automated task was the routine layer that allowed workers to build experience, removing it can narrow the path into the occupation. The remaining work may be more valuable, but fewer people can reach it. A technology can therefore change the ladder into a profession before it changes the number of people at the top.

Expertise is also different from formal education. The evidence discussed by the IMF associates task-based expertise with hourly earnings and estimates that expertise alone accounts for roughly one-third of observed wage variation across occupations, even after controlling for education. That does not mean education is unimportant. It means that a transition can change wage prospects through practical capability and task content in ways that a credential category does not fully capture.

This distinction helps explain why the same AI investment can create both opportunity and pressure. A firm that automates routine parts of a process may need fewer people for the old workflow but more workers able to supervise, interpret and integrate the new one. If those complementary workers are scarce, their wages may rise.