The Winners and Losers of the AI Revolution | Tyler Cowen [ARC 2026]

Overview

Tyler Cowen argues that even with “strong AI” arriving, it won’t eliminate jobs so much as transform them—automating the most tedious, repetitive parts first while creating new categories of work. He frames the transition as a “good news/bad news” situation: work won’t disappear, but people and economies must adapt quickly.

Main Points and Analysis

Jobs evolve rather than vanish

Cowen compares the AI era to earlier historical periods (notably pre–Industrial Revolution England), where new technologies altered job structures. In those times, people couldn’t easily predict what new job titles would emerge.

Big new job growth areas in an AI-drenched world

• Energy / infrastructure buildout

  • Every major country must expand energy systems.
  • This creates executive and managerial opportunities as careers shift location and sector (he notes examples like moving from finance to energy leadership).

• UK energy risk

  • Cowen is particularly concerned about the UK, citing higher electricity costs (about “four times” Texas).
  • He also mentions that OpenAI’s data-center plans in the UK were reportedly called off—suggesting some job problems may stem from policy and infrastructure lag.

• Labor mobility as key policy lever

  • His central policy recommendation is improving mobility across regions and sectors.
  • Countries with low mobility will struggle because AI-driven job reallocation requires people to move and retrain; countries with higher mobility will “thrive.”

• Retraining everyone

  • The largest adjustment is that nearly everyone must learn to use AI tools.
  • Starting sooner can improve productivity rather than causing alienation and confusion.

• Data gathering and data preparation

  • AI will need more data from governments, universities, companies, and archives that hasn’t been fed into models yet.
  • Transforming and structuring this data becomes a major job sector.

• Running experiments (especially biomedical)

  • Demand for experimentation could rise dramatically (he suggests ~10x).
  • Even with AI-generated hypotheses, real-world testing and regulatory-approved trials remain necessary.

• Entertainment and cultural “front-people”

  • Even if AI produces music, art, or jokes, Cowen believes humans will remain central as emotionally resonant performers/creators.
  • He highlights charisma, memorability, and “front” roles.

• Artisanal construction and custom craftsmanship

  • As mass production becomes cheaper, demand grows for unique, crafted, original goods—an anti-“AI slop” trend.

• “Weird” service jobs

  • He expects growth in personal and human-facing services (e.g., assistants, greeters).
  • Affluent people may maintain retinues of helpers, including AI agents—yet still requiring human roles.

• Caring for the elderly

  • Continued and expanded demand is expected, framed as a long-known trend unlikely to reverse.

How Jobs Will Change in Character

• Hard-to-describe work becomes more valuable

  • AI handles jobs that are easy to define.
  • Human value increases in tasks involving messy, social, varied, or relational elements.

• Company size shrinks

  • AI makes it easier for small teams to accomplish a lot.
  • More startups and small firms can grow quickly, often outsourcing or hiring later as projects scale.

• Companies become “weirder” and more unique

  • With smaller optimal firm sizes, businesses may feel more personal and less impersonal.
  • This could produce highly specialized, memorable companies (he cites Midjourney as an example).

Who Wins and Who Loses (“AI maniacs” vs. rule-following elites)

• Biggest winners: “AI maniacs”

  • These are people who aggressively learn how to use AI and orchestrate it with business processes/agents.
  • They may not build models themselves, but can coordinate their use effectively.
  • Cowen suggests this can enable outsized success (with potentially very high upside but also a high failure rate).

• Relative decline: rule-following elites

  • He predicts a decline for a politically influential class who follow established rules—well-trained, conscientious workers expecting a traditional path to partnership and high salaries.
  • Unless they become “AI natives,” their advantage may erode.
  • They likely won’t necessarily starve, but may lose relative ground.

Political risk

Cowen warns that shifting income/class structure could become politically destabilizing because the educated elite losing relative edge remains highly influential.

A Paradox About Leisure Time

In the long run, AI should generate leisure through productivity and wealth, but in the short run (next ~5 years) people should expect to work harder—because they must learn and adapt to AI. The “leisure dividend” arrives later than many expect.

Conclusion / Framing

Cowen’s thesis is that AI will reorder labor markets rather than erase employment. Countries and individuals that invest in retraining, mobility, energy/infrastructure, and the ability to run experiments and manage data will benefit most, while those that resist adaptation will face worse outcomes.

Presenters or Contributors

• Tyler Cowen