AI Will Recut Jobs Before It Replaces Them

“Which jobs will AI replace?” uses the wrong unit. Jobs bundle research, communication, judgment, entry, coordination, and accountability. Those tasks have different automation profiles. AI changes the bundle first; organizations later decide whether to shrink, expand, or rename the role.

Exposure also differs within an occupation. Junior staff may spend more time collecting and organizing information, while senior staff emphasize judgment and trust. Automating junior work can remove not just toil but the apprenticeship through which expertise was acquired.

New tasks appear around input and acceptance: defining goals, preparing context, setting permissions, validating output, handling exceptions, and owning the decision. Activities once scattered across jobs may become explicit operations and assurance roles.

Organizations may flatten initially, then add a new middle layer. A few experienced workers with agents can produce more, but scale creates a need to manage models, tools, data, and quality. The one-person-company thesis holds only where liability and coordination remain bounded.

Pay will not divide simply between people who use AI and those who do not. Greater value accrues to people who translate ambiguous business problems into executable work, recognize failure modes, and embed output in real decisions. Tool fluency diffuses faster than accountability.

Transition costs are concentrated. Firms can capture automation gains immediately, while workers need time to learn new tasks. Without training, internal mobility, and social support, recomposition appears as instability, wage pressure, and disappearing entry routes.

Managers should measure tasks before headcount: time, error cost, necessary human judgment, and ownership after automation. Rebuilding the workflow before setting staffing targets is more reliable than choosing a reduction and then searching for an AI justification.

The first labor shock is not an orderly set of vanished occupations. It is a recutting of work boundaries. The consequential choice is whether the new bundles create better work or merely transfer risk and supervision to fewer people.