AI Exchanges: AI’s Impact on Employment

Episode Summary

Executive Summary: The episode examines whether AI is already affecting jobs and concludes that, so far, labor-market effects are modest because adoption remains early. Goldman Sachs research finds AI use in regular production at about 9% of U.S. companies, with stronger uptake at large firms and in tech, finance, education, and business services. The guests argue the near-term risk is transitional displacement, especially for younger and junior workers, while long-run effects depend heavily on adoption speed and whether AI evolves toward AGI.

Main Topics: AI adoption is still early and macro labor effects are limited (Priority: 5/5): Joseph Briggs says regular-production AI adoption is only around 9% of U.S. companies, so broad labor-market impacts are not yet visible in aggregate indicators like unemployment, layoff rates, hours worked, or wages. Tech is the clearest early casualty and signal (Priority: 5/5): The technology sector shows the strongest early headwinds, with hiring below trend and more AI-related job postings, suggesting both productivity gains and displacement pressure in software-heavy roles. Younger and junior workers are seeing the first signs of strain (Priority: 5/5): The discussion highlights lower hiring for recent graduates and a larger unemployment increase among tech workers ages 20-30, though some of this reflects a broader low-hiring labor market rather than AI alone. Displacement may be substantial, but likely transitional (Priority: 5/5): Briggs estimates full-adoption displacement at 6%-7%, emphasizing that the key issue is whether job losses occur over a few years or a decade-plus, which determines how disruptive the unemployment spike would be. Jobs most resilient to AI share common traits (Priority: 4/5): Roles with high human interaction, high stakes, and nonrepetitive work are less exposed; examples include teachers, clergy, pharmacists, medical care providers, CEOs, and door-to-door salespeople. Adoption speed and economic cycles could magnify disruption (Priority: 4/5): The panel warns that recessions could accelerate automation-driven layoffs, concentrating displacement into a shorter period and raising unemployment more sharply than in a gradual transition. AGI and general-purpose technology remain the long-tail uncertainty (Priority: 4/5): The speakers note that their forecasts exclude AGI. They compare AI more to electricity and telephony than to internet/cloud alone, arguing that general-purpose technologies take time to reshape the economy.

Key Arguments: AI’s labor-market impact is currently small because only a minority of firms are using it in regular production. Large firms are adopting AI faster than small firms, likely because they have the expertise to build internal tools; smaller firms are waiting for plug-and-play solutions. Tech is both the most exposed sector and a possible canary in the coal mine for broader employment effects. AI-related hiring is rising, but these roles represent a small slice of the economy relative to the jobs potentially being automated. Recent college graduates and young tech workers are showing weaker hiring and higher unemployment, indicating early pressure on entry-level pipelines. The most likely near-term outcome is transitional unemployment rather than permanent structural unemployment. The size of any unemployment shock depends less on the displacement estimate itself than on how quickly adoption and application build-out occur. Recessions may intensify automation because firms often target routine occupations they expect to automate anyway. Jobs that rely on human judgment, social interaction, and high-stakes decision-making are relatively resilient to automation. History suggests technology has historically created more jobs than it destroyed, but AI could differ if it leads to AGI or much faster innovation. The research framework uses general-purpose technology analogs such as electricity and the IT revolution, while acknowledging limited historical data points. A hybrid workplace with humans and AI agents may emerge, creating new management and apprenticeship challenges.

Data Points: U.S. companies using AI in regular production: 9% - Current estimated adoption rate cited by Joseph Briggs; defined as using AI for regular production of goods and services over the last two weeks. AI adoption among large companies: mid-to-high teens - Firms with more than 250 workers are adopting faster because they can develop in-house tools. AI-related job postings growth: 25% to 50% - Job postings mentioning AI have increased relative to other postings in the latest adoption tracker. Tech unemployment increase for young workers: about 3 percentage points - Unemployment for tech workers ages 20-30 has risen since the start of the year, more than for tech overall or other young workers. Estimated transitional displacement from full AI adoption: 6% to 7% - Briggs’s estimate of workers who may lose jobs due to AI-related automation during the transition. Unemployment effect after a 1 percentage point productivity boost: around 30 basis points - Historical estimate: technological productivity gains typically raise unemployment over the next year, with no effect after two years. Potential unemployment-rate boost if displacement occurs over 1-3 years: 2% to 2.5% - Fast AI adoption could concentrate displacement and create a significant macro shock. Potential unemployment-rate boost if transition takes 10-15 years: about 0.5 percentage point or less - Slower adoption would spread displacement over time and make the labor-market impact more manageable. Long-run job growth driven by technology: 85% over the last 85 years - Used to argue that technology historically creates new positions even as it displaces old ones.

Pivotal Quotes: "our AI productivity growth forecasts have always assumed a 6% to 7% displacement rate" — Joseph Briggs: Briggs explains the size of expected transitional displacement under full AI adoption. "If we look at the overall labor market data so far, it looks pretty small to me." — Joseph Briggs: He summarizes the current macro evidence on AI’s labor-market impact as limited. "the rise of agents, their utility in the enterprise, we have to climb a hill of maturity there before we see those really emerge as being effective." — George Lee: Lee argues that AI agents will matter more as the technology matures and becomes operationally useful in enterprises.

Implications: AI’s job impact is likely to be gradual but uneven: entry-level and routine roles may be pressured first, especially in tech, while resilient jobs and hybrid human-AI workflows could expand as adoption matures.