History's most reliable lesson about technological change is that the jobs created by a new technology are almost never the ones people predict, and almost always more numerous than the catastrophists feared. In 1800, no one predicted "railroad conductor," "telegraph operator," or "steam engine mechanic." In 1950, no one predicted "software developer," "database administrator," or "UX designer." In 1990, no one predicted "social media manager," "app store developer," or "data scientist." The new jobs consistently emerge in three waves: first, jobs building and maintaining the new technology; second, jobs applying it in existing sectors; third — largest and most surprising — jobs made possible only because the technology exists and has lowered some cost so dramatically that entirely new industries become viable.
The AI wave is already generating its first-wave jobs: prompt engineers, AI trainers, model evaluators, AI safety researchers, machine learning infrastructure engineers, synthetic data producers, AI auditors, and ethics reviewers. These roles did not exist in meaningful numbers five years ago. They represent direct employment in the construction and governance of the new technological layer.
The second wave is application work. Every industry that deploys AI needs specialists who understand both the domain and the AI system — medical AI deployment managers, legal technology specialists, AI-assisted financial advisors, AI-enhanced architects, precision agriculture analysts. These hybrid roles combine deep domain expertise with technological fluency. They are currently underserved and will grow rapidly as organizations navigate the integration challenge. The person who can translate between what an AI system can do and what a domain actually needs is consistently underpaid and undersupplied.
The third wave is the most speculative and most historically large. When a new technology collapses a major cost category, it unlocks demand that previously could not exist. Steam power made oceanic freight affordable, which expanded global trade, which created entirely new occupations in logistics, customs, insurance, and international finance. The internet made communication nearly free, which created social media, search advertising, e-commerce, content creation, and platform governance — all industries that existed at vanishingly small scale before 1995. The question is: what cost is AI collapsing, and what new industries will that unlock?
The most plausible candidates are in personalized services. When expert cognitive labor was expensive, personalized tutoring, personalized healthcare, personalized legal advice, and personalized financial planning were luxury goods. AI makes customized cognitive assistance far cheaper, potentially extending these services to billions of people who previously had access only to generic, mass-market versions. The human jobs that emerge are not the AI itself but the orchestration layer: the tutor who curates and delivers AI-assisted personalized education, the community health worker who interprets AI-generated health assessments, the legal navigator who helps working-class clients use AI-assisted document preparation. Democratized access to expert tools creates a new class of paraprofessional who amplifies their own expertise with AI and serves populations previously underserved by professionals.
Another candidate is the experience economy at scale. As AI handles information processing, human comparative advantage shifts toward activities that are intrinsically valuable as experiences: live performance, embodied learning, in-person community, therapeutic encounter, artisanal production. The job category of "experience designer" — someone who architects meaningful human interactions at the intersection of physical and digital — is growing and will continue to grow.
The challenge is that the third wave takes time. Between wave one and wave three, there is a displacement trough. The people who lose jobs in wave one are not automatically the people who get jobs in wave three. The transition requires retraining, geographic mobility, institutional support, and sometimes generational replacement. The aggregate outcome is positive by most historical measures; the distributional outcome depends on policy.