Does new AI automate your job away, or hand you new, higher-value work to do instead?

This explores whether some kinds of AI tend to create work while others take it away. The short answer from the collection: the split depends less on the technology than on what it is built to do. The same model can replace a worker or give them new work, and the deciding choice is made by the firm deploying it. Acemoglu, Autor and Johnson sort AI by its effect on labor. They argue that AI which takes over expert tasks pays firms more than AI which creates new tasks for people. So firms build too little of the pro-worker kind, even when society would be better off with more of it Why do firms build automating AI instead of pro-worker AI?. Brynjolfsson adds that the choice is about power as well as money. Automation moves bargaining power to whoever owns the system, and that can lock in a world where workers have no leverage to change course Does automating labor concentrate power more than augmenting it?.

What happens in practice is mostly reshaping, not mass replacement. MIT studied more than 20 companies and found that generative AI rolled out slowly over three years. Headcount held steady. Workers moved from doing the task to supervising it, the same shift aviation and factory automation went through decades ago Does generative AI adoption actually eliminate jobs or reshape them?. Payroll data through mid-2026 shows no economy-wide job losses. There is one sharp exception: young workers in AI-exposed jobs are hired about 19% less often than their peers, while experienced workers show no such gap Is generative AI displacing workers at economy-wide scale?. AI isn't firing people so much as closing the entry door.

One less obvious factor is how AI's reach is spread across a job. When AI touches only a few tasks in a role, people move to the tasks it can't do, and net job effects stay small. When it touches many tasks at once, there is nowhere left to move Does concentrated AI exposure enable workers to adapt and reallocate?. Firms also differ widely. Companies with heavy AI exposure replace freelance workers faster and more cheaply than others, which points to in-house AI skill building on itself rather than everyone adopting at the same pace Do firms substitute labor for AI at different rates?. Where workers have actually handed tasks to AI, it is mostly information work, and it follows what the technology can do rather than older predictions that routine jobs would go first Where have workers actually delegated tasks to AI?.

On the augmentation side, the gains are real but limited. In a study of 758 consultants, GPT-4 helped weaker performers the most and narrowed the skill gap. That held only when AI use was required; when people could choose, the effect could reverse Does AI help lower-skilled workers catch up to top performers?. Productivity gains also appear mainly when workers use AI on skills they already have. When they use it to learn something new, the gains disappear and learning suffers When does AI actually boost worker productivity?. Interviews with knowledge workers found four outcomes rather than a simple upskill-or-deskill split: skills can develop, stay the same, erode, or change in value How does generative AI actually change worker skills?. Workers seem to sense this. In Anthropic's survey of 81,000 users, fear of losing a job was highest both among people AI slowed down and among people it sped up the most Does AI productivity gain always ease job displacement fears?.

A gap worth knowing about: the collection has little direct evidence of AI creating new job categories. Most of what it measures is displacement, reshaping, or productivity within existing roles. The deeper stakes may also go beyond paychecks. One line of argument holds that institutions stay aligned with human interests partly because they depend on human workers who care about outcomes. If that dependence fades bit by bit, society could slowly lose influence over its own systems Does incremental AI replacement erode human influence over society?.