AI and Jobs

Any honest conversation about AI and jobs has to start by acknowledging that the honest answer is: we do not know exactly what will happen. Economic forecasters have a poor track record predicting how new technologies reshape labor markets. What we can do is look carefully at the patterns from history, the current evidence, and the mechanisms at work — and think clearly about what they mean for the people who will be working in the economy you are about to enter.

What History Tells Us

Automation anxiety is not new. In the 1960s, economists warned that computing would create mass unemployment. In the 1980s, robots on assembly lines were supposed to eliminate manufacturing work. Neither prediction came true in the way predicted — but both caused genuine disruption for specific workers and communities, even as new jobs emerged elsewhere. The pattern from most technological transitions is roughly this: technology eliminates some tasks (specific things humans do), which sometimes eliminates jobs, but it also creates new demand for goods and services, lowers prices, and creates new categories of work that did not previously exist. Call center work barely existed before the telephone. Web developers did not exist before the internet. Social media managers did not exist before social media. The key nuance: the people who lost jobs in textile mills in the nineteenth century were not automatically the people who got new jobs in railways. Transitions create winners and losers, and the losers are often specific communities — often already disadvantaged — who bear concentrated costs while broader society gets diffuse benefits.

Current evidence suggests AI is particularly effective at tasks that are: routine and rule-based (data entry, basic customer service, document classification), based on pattern recognition in structured data (image classification, fraud detection, quality control), and involve generating standard text from templates (boilerplate legal documents, form letters, routine reporting). Tasks that remain harder to automate include: those requiring physical dexterity in unpredictable environments (plumbing, surgery, childcare), novel problem-solving without a large base of prior examples, work requiring trust, empathy, and sustained human relationships, and work requiring ethical judgment in complex, novel situations. This suggests that high-skill, high-education jobs with creative and interpersonal components may be less vulnerable in the near term — but also that some white-collar knowledge work (legal research, basic coding, routine analysis) is more automatable than previously assumed. Generative AI has changed this calculation significantly since 2022.

New Roles That AI Creates

Every wave of automation has created new work. AI is already creating categories of employment that barely existed a few years ago. AI trainers and annotators label data, write feedback for AI outputs, and do the human quality-control work that makes AI systems reliable. As of 2024, this is a significant global industry. Prompt engineers design effective inputs for AI systems — work that requires deep knowledge of both the domain and the AI's behavior. AI safety and alignment researchers study how to make AI systems behave as intended — a research field that is growing rapidly. AI auditors evaluate AI systems for fairness, accuracy, and compliance — a role that barely existed five years ago and is now required by emerging regulations in the EU and US. The question of who benefits from new AI-created jobs is as important as whether they exist. If AI eliminates jobs that are accessible to people without college degrees and creates jobs that require advanced technical skills, the net effect on economic inequality could be significant even if the total number of jobs stays the same.