Introduction: Rising Anxiety in the AI Era
Concerns about artificial intelligence and jobs are widespread. Moreover, many workers fear automation will reduce opportunities and wages.
Simultaneously, inequality and labor’s income share have worsened over recent decades. Therefore, technological direction now carries profound economic and social consequences.
Defining Pro-Worker AI
Pro-worker technologies expand human capabilities and increase the value of expertise. Importantly, not all productivity-enhancing technologies meet this demanding standard.
Some technologies automate tasks, while others create new human work. Thus, evaluating AI requires distinguishing how it reshapes labor demand and expertise.
Five Types of Technological Change
Technological change falls into five broad categories. First, labor-argumenting technologies increase worker productivity in existing tasks.
Second, capital-augmenting technologies improve machines performing established tasks.
Third, automation technologies substitute machines for human-performed tasks.
Fourth, expertise-leveling technologies enable less-workers to perform advanced tasks.
Fifth, new task-creating technologies generate demand for entirely new human expertise.
Automation: Benefits and Trade-Offs
Automation raises productivity by reducing costs and replacing labor. However, it often commodifies specialized expertise and reduces labor demand.
Consequently, automation can lower labor’s share of income. Although consumers benefit from lower prices, displaced workers bear concentrated costs.
New Task Creation: Unambiguous Gains
In contrast, new task-creating technologies expand what humans can productively do. They increase both the quantity and variety of expertise-based work.
Additionally, they raise labor demand and strengthen labor’s share. Therefore, they are unambiguously pro-worker. Historically, new occupations have offset automation’s erosion of expertise.
Expertise-Leveling: Mixed Effects
Expertise-leveling technologies broaden access to advanced tasks. For example, tools can help less-experienced workers perform complex functions.
Yet incumbents may face intensified competition. Thus, gains for some workers may coincide with losses for others.
AI’s Collaborative Potential
AI differs from traditional computing because it handles unstructured information and judgment. Rather than following fixed rules, AI learns patterns and synthesizes context.
Accordingly, AI can collaborate with workers instead of replacing them. Collaboration enables workers to perform sophisticated tasks and acquire expertise faster. Importantly, collaborative tools need not be flawless to add value.
Why Automation Dominates
Despite AI’s collaborative potential, automation remains dominant. First, forms often perceive higher returns from replacing labor.
Second, technological development follows established automation paths.
Third, an ideological focus on artificial general intelligence prioritizes human substitution.
Therefore, pro-worker AI is underdeveloped relative to its potential.
Market Failures Limiting Pro-Worker AI
Misaligned incentives discourage investment in expertise-enhancing systems. Path dependence reinforces automation-oriented research trajectories.
Moreover, markets may undervalue social benefits of expanded worker capabilities. Consequently, private investment alone may not yield optimal technological direction.
Policy Directions to Shift Incentives
Public investment in health care and education can encourage collaborative AI. Government capacity in AI expertise can shape procurement and standards.
Targeted grants can support task-creating innovation. Furthermore, tax reforms may rebalance incentives away from pure automation.
Antitrust enforcement can promote competition and innovation diversity. Protections against expertise theft can safeguard worker knowledge. Finally, worker voice can influence design and deployment decisions.
Conclusion: Choosing AI’s Direction
AI can either commodify expertise or expand it. Although automation increases productivity, it often deepens inequality.
By contrast, new task creation strengthens labor demand and income share. Therefore, shaping AI toward collaboration and expertise expansion is both feasible and desirable.
Source:
Acemoglu, D., Autor, D., & Johnson, S. (2026, February). Building pro-worker artificial intelligence. The Hamilton Project, Brookings Institution.