Bank of America27 min
Why we believe AI reshapes work more so than it reduces overall payrolls
Benson Wu, Nick Stenner, TJ Thornton
Earnings Call, Interview, Conference Presentation
Core Thesis on Labor Impact
- BofA Global Research rejects "apocalyptic" job destruction narratives, arguing AI will primarily transform specific tasks rather than eliminate entire occupations at scale.
- While approximately 25% of global jobs are exposed to generative AI (per ILO data), only 2.3% have high automation potential, whereas 13% offer significant augmentation potential.
- Roles facing shrinkage include collaborative coding support, customer service, administrative work, data entry, and specific ICT functions.
- The primary shift involves reallocating human effort from routine tasks (drafting, searching, processing) to higher-value activities like interpretation, decision-making, communication, and strategy.
Emerging and Evolving Job Categories
- AI Specialists: New roles include prompt engineers, AI trainers, model evaluators, and positions requiring ethical judgment and domain expertise.
- Hybrid Professionals: Existing roles will expand to integrate AI tools, such as research analysts focusing on interpreting AI-generated data or lawyers utilizing AI for document review while shifting to legal strategy.
- Human-Centered Services: Sectors like healthcare, elderly care, education, and personal services are expected to grow due to higher real incomes and shifting consumer demand toward trust-based interactions.
- Entrepreneurship: Agentic AI may lower barriers to entry, enabling "one-person companies" and fostering new small business formation.
- Historical precedent shows that 60% of current US jobs did not exist in the 1940s, suggesting new roles will emerge endogenously as productivity increases.
Historical Parallels and Adaptation Capacity
- Past General Purpose Technologies (Industrial Revolution, Internet, Excel) displaced tasks but ultimately raised productivity, lowered costs, and created unimagined jobs.
- Advanced economies (Europe, Americas) face the highest exposure rates (approx. 33% of jobs) compared to low-income economies (approx. 11%), yet are better positioned to adapt.
- Adaptation drivers in advanced economies include higher human capital, advanced education systems, deep digital infrastructure, and flexible labor markets.
- Aging demographics in Europe, North Asia, and North America may benefit from AI offsetting labor shortages by boosting output per worker.
Client Feedback and Contested Views
- Investors generally agree on high exposure for white-collar cognitive jobs but debate the interpretation of risks associated with "agentic AI" (autonomous action capabilities emerging in late 2024/2025).
- Concerns exist that capital owners may capture disproportionate wealth, potentially limiting the consumer base for new service-sector jobs.
- The research team anticipates government intervention to ensure even distribution of productivity gains and facilitate labor transitions.
Wages, Inequality, and Labor Distribution
- Aggregate wage data from the post-2022 period shows no direct, broad-based decline in wages linked to AI exposure across 200+ US occupations.
- Dispersion risks exist: experienced workers utilizing AI effectively may see wage growth, while demand for entry-level routine tasks may shrink.
- A primary risk is the disproportionate flow of productivity gains to capital owners and firms with strong data/infrastructure, potentially increasing the capital share of income.
- Policy levers for mitigation include training programs, labor mobility support, wage insurance, and tax design reforms.
Youth Employment Dynamics
- Youth unemployment (ages 22-25) in AI-exposed occupations has declined significantly, but this cannot be solely attributed to AI.
- Structural factors, such as a doubling of new graduates in China over the last decade and slowed labor demand in the property sector, are significant drivers of weak youth employment.
- AI may specifically reduce demand for entry-level routine work that traditionally served as a training ground for early-career professionals.
Macroeconomic Implications and Interest Rates
- AI-driven productivity is theoretically disinflationary over the long term by lowering unit labor costs and expanding supply capacity.
- Central banks may gain room to tolerate temporary inflation spikes if productivity gains materialize, though immediate effects on rates may be limited.
- Productivity growth could paradoxically push up the neutral real rate by increasing the return on capital, creating a complex policy trade-off.
- Transition costs, including potential fiscal spending on labor market adjustments, could increase bond supply and pressure long-term rates.
- Gains are likely to be uneven, with early productivity spikes concentrated in leading firms rather than the aggregate economy.