S6 Ep48: AI’s impact on jobs

Episode Summary

Executive Summary: John Van Reenen argues that AI’s effects on productivity and employment remain uncertain, and that it is unlikely to reverse slow productivity growth quickly. Unlike industrial robots, AI can affect a broad range of cognitive tasks. It may strengthen dominant firms and reshape job hierarchies, while creating new work as it displaces old tasks. The best response, he says, is not to block AI but to support retraining, flexible education, and compensation for those who lose out.

Main Topics: Uncertainty and productivity prospects: AI is at an early, fast-changing stage, making its economic effects difficult to forecast. Van Reenen sees potential for productivity gains but cautions that deployment and organizational change take time, so AI is unlikely to be a near-term fix for weak growth. AI versus industrial robots: Robots have mainly automated repetitive manual work, while AI can affect cognitive tasks across many industries, including coding, legal research, and writing. This makes AI’s potential reach across jobs much broader. Superstar firms, data, and inequality: AI software may lower barriers for smaller businesses, but large firms’ access to extensive data and highly skilled workers can reinforce their advantages. The benefits may also accrue disproportionately to AI creators and owners. Changes to jobs and workplace hierarchies: Routine coding and information-processing tasks may face declining demand. Remaining workers may need broader skills to check AI outputs, coordinate work, and make decisions, while some organizational layers may shrink and senior roles retain greater influence. Creative destruction and a just transition: Technology destroys some jobs while creating others, and Van Reenen does not expect society as a whole to run out of work. He argues that displaced workers need reskilling, pathways into new roles, and income support; simply blocking adoption is impractical. Education and assessment in the AI era: Education should become more flexible and support lifelong learning rather than treating a degree as one-time preparation. AI also challenges conventional take-home assessments, requiring new ways to evaluate learning and skills.

Key Arguments: The productivity gains from a general-purpose technology depend on complementary changes in how firms operate; invention alone does not immediately raise output. Electricity took decades to transform productivity, suggesting AI may also have a long adoption and reorganization period. AI is potentially more pervasive than factory robots because it can assist with or automate cognitive tasks found across many occupations, not only repetitive manual work. AI may reinforce large firms because data improves algorithms and personalization, and major companies can attract scarce AI talent. Software’s copyability could help entrants, but does not eliminate incumbents’ data and scale advantages. AI is likely to change the composition and organization of work rather than eliminate all employment: some tasks and jobs will disappear, while new roles and demands emerge. Workers who remain may take on broader responsibilities and oversight. Effective policy should enable workers to move between jobs through retraining and flexible labor-market institutions, while using social security and redistribution to compensate people who bear the costs of change. Education systems should provide learning throughout people’s working lives and adapt assessment to a world where students can use generative AI; merely restoring closed-book, in-person testing is not a sufficient general solution.

Data Points: UK annual productivity growth before the global financial crisis: About 2.5% - Van Reenen contrasts the roughly twelve years before the 2008–2009 crisis with the period afterward. UK annual productivity growth after the global financial crisis: About 0.5% - The approximate rate in the dozen years after the crisis, before COVID, indicating a substantial slowdown. Time from electricity’s arrival to major productivity effects: About 20–40 years - Electricity emerged around the 1880s, but its transformative effects required later changes such as electrified factories and production lines. Time period of weak productivity growth since the financial crisis: About 15 years - Van Reenen describes the period after the 2008–2009 global financial crisis as a prolonged productivity downturn. Audience for a virtual PhD lecture: Thousands of people - Van Reenen cites a lecture on climate change and innovation to illustrate how technology can expand educational reach.

Pivotal Quotes: "I certainly, if I was like the British Chancellor of the Exchequer, I wouldn't be banking on it to rescue us over the next five years." — John Van Reenen: His caution that AI is not likely to be an immediate solution to the productivity slowdown. "technology isn't destiny." — John Van Reenen: His point that society’s rules and policy choices shape who receives the economic gains from AI. "The best way to do it is through this flexi security idea, which is to create a society where people can get retrained and move jobs and new roles to get new tasks." — John Van Reenen: His proposed approach to managing job disruption through retraining and worker mobility.

Implications: AI’s gains will depend on adoption, workplace redesign, and access to data—not just the technology itself. Workers and employers should prepare for changing tasks through lifelong learning, while policymakers need stronger routes to retraining and support so productivity gains do not deepen inequality.