How Will AI Impact the Labor Market?

Goldman Sachs Research Forecasts

• Near-term displacement: AI is currently creating a 10,000 to 15,000 monthly drag on job growth in specific sectors including tech, management consulting, and graphic design.
• Long-term baseline scenario: Full AI adoption could yield a 15% productivity uplift, potentially reallocating approximately 9% of the U.S. workforce (approx. 15 million workers) over a 10-year period.
• Unemployment impact: Despite significant reallocation, the forecast expects the unemployment rate to rise by less than one percentage point in any single given year.
• Historical precedent: Over the last 80 years, roughly 85% of job growth has been driven by technological creation of new positions.
• Reabsorption capacity: A mere 5% acceleration in the pace of new job creation would be sufficient to reabsorb workers displaced by AI-driven automation.
• Churn dynamics: The U.S. labor market currently destroys approximately 29 million jobs annually while creating 30 million, a pattern expected to continue with AI.

MIT Perspective: Adoption Barriers and Task Dynamics

• Capability vs. Reality: AI task capability growth does not directly equate to job displacement; adoption requires successful integration of data access, system building, and economic viability.
• Adoption asymmetry: Large enterprises are expected to automate tasks before small businesses due to resource availability and scale.
• Partial automation effects: Unlike previous automation waves focusing on routine tasks, AI often automates subsets of jobs, leading to varied outcomes based on which tasks are affected.
• Expertise displacement scenario: If AI automates the "expert" core of a role while leaving mundane tasks to humans (e.g., proofreaders losing spell-check duties), the number of workers may increase, but wages could decrease.
• Wage and employment inversion: If AI automates "inexpert" tasks (e.g., taxi drivers losing route knowledge to GPS), remaining experts may see wage increases, but total employment for that role may shrink.
• Implementation friction: AI's broader scope and lack of 100% reliability (hallucinations) make it superior as a human tool rather than a fully autonomous process component.
• Adoption timeline: The labor market impact will follow a "rising tide" pattern of gradual depth increase rather than a "crashing wave" of sudden mass displacement, allowing for managed adaptation.

MIT Perspective: Limitations and Broader Impacts

• Job longevity: While some jobs will be largely automated and disappear, the net balance between job destruction and new task creation remains an open question for AI specifically.
• Reliability constraints: Unlike deterministic tools like Excel, AI's probabilistic nature requires human oversight, limiting its use as a standalone "set and forget" system.
• Organizational scaling: AI tools may allow firms to expand faster; while individual roles disappear, the firm's expansion could lead to net hiring.
• Broader scope: AI's ability to handle diverse tasks (from math to dinner planning) makes it more powerful but also harder to integrate reliably than narrower technologies.

MIT Perspective: Future Trajectories and Inequality

• Near-term outlook: AI investment in replacement over complementation could result in a net negative labor impact of 2% to 4% over the next five years, with limited layoffs expected in 2027.
• Current limitations: Widespread displacement is hindered by a lack of reliable, easy-to-use applications built on foundation models for general enterprise adoption.
• Sector exception: Software engineering and coding are immediate outliers where impact is visible due to existing model capabilities and high user expertise in prompting.
• Vulnerable roles: Cognitive, routine tasks involving limited social interaction or judgment (e.g., customer service, back-office work) affect approximately 8–9 million U.S. workers.
• Long-term uncertainty: Net job losses could accelerate over a 10–15 year horizon if AI investment continues to prioritize worker replacement.
• Robotics wild card: The integration of AI with robotics could vastly expand the scope of impacted work, as physical jobs constitute roughly 50% of the U.S. economy.
• Social interaction barriers: Jobs requiring social interaction need both improved AI social capabilities and shifts in human consumer preferences to be viable for automation.
• Inequality projection: Labor income inequality is expected to increase as AI replaces lower-to-mid-level cognitive roles; replacing high-paid managers is deemed less likely and could theoretically reduce inequality.
• Historical parallel: The period since the 1970s has seen job creation fail to match destruction, particularly for workers without college degrees; a similar dynamic in cognitive jobs could mirror this trend.

General Consensus and Disagreements

• Disagreement on magnitude: Joseph Briggs (Goldman Sachs) and Neil Thompson (MIT) anticipate significant reallocation but argue job creation will offset losses, whereas Darren Acemoglu (MIT) warns of potential net negative labor impact in the short-to-medium term.
• Agreement on inevitability: All experts agree that AI will displace a significant number of workers, though the speed and scale depend on the path of investment (complementary vs. replacement).
• Agreement on timeline: The transition is expected to be gradual and uneven, driven by the pace of reliable application development rather than raw model capability.
• Forward-looking warning: Acemoglu highlights that the specific trajectory of AI development (complementary vs. replacement) is the critical variable determining long-term employment levels.
• Recorded date: These insights were recorded in June 2026.