Word count: 539
Artificial intelligence (AI) doesn’t just automate work—it may create cascading inequalities that accumulate across design, implementation, and use. A comprehensive review of 335 studies reveals how small differences at one stage amplify into wider gaps in evaluation, wages, and workplace relationships over time.
Why it matters:
AI reshapes how work is organized, evaluated, rewarded, creating multiple points where inequality emerges and compounds. Small disparities introduced during AI design can later shape evaluation norms, productivity expectations, and team dynamics, causing what the researchers call “inequality cascades.”
AI-driven task restructuring can widen wage and mobility gaps as some workers benefit disproportionately from augmentation while others face deskilling or displacement.
As organizations deploy AI across functions and workflows, inequality risks become structural rather than isolated. Leaders need to evaluate not just how AI performs but its social and relational consequences within the workplace.
How we know:
The authors conducted an integrative review of 335 peer-reviewed articles, using Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) procedures across management, economics, sociology, information systems, and public policy. They identified four exisiting perspectives in the literature—normative, cognitive, structural, and relational—and showed how these mechanisms interact across the AI lifecycle to produce inequalities.
What the researchers found:
- Encoded inequality arises when AI systems inherit biased data or reflect design choices that prioritize efficiency over fairness.
- Evaluative inequality occurs when workers’ perceptions of AI (trust, judgments, and interpretation) shape performance outcomes, credibility, and access to opportunity.
- Wage inequality emerges as AI redistributes tasks and productivity gains, often benefiting higher-skill or AI-literate workers more than others.
- Relational inequality occurs when AI alters authority, autonomy, expertise, and coordination within teams.
- These four types of inequality and amplify interact across the AI lifecycle, from design through implementation to use, creating cumulative inequality cascades.
- Organizational choices, worker perceptions, and system feedback loops continuously shape these inequalities, meaning technical solutions alone cannot address them.
What this means:
- For executives: Treat AI adoption as an organizational transformation, not a technical upgrade; assess inequality risks across the full AI lifecycle and align AI strategy with long-term capability building and inclusion.
- For HR, learning, and governance leaders: Monitor how AI changes skill demands, evaluation criteria, and internal mobility; invest in training programs that support human-AI collaboration and equitable skill development. Combine technical bias checks with contextual and cross-functional insight, and continuously monitor how AI affects real decisions and workflows.
- For managers & team leads: Help teams navigate automation bias, communicate clearly about how AI informs decisions, and ensure that workers retain meaningful autonomy.
Now what:
- Assess AI systems for encoded, evaluative, wage, and relational inequalities and identify where early disparities may trigger broader inequality cascades across the organization.
- Examine AI across the full design–implementation–use lifecycle to understand how assumptions, user reactions, and workflow changes may compound into persistent disparities.
- Strengthen worker and manager AI literacy so employees can interpret AI recommendations accurately, avoid automation bias, and engage with AI systems in ways that support fair outcomes.
- Monitor how AI reshapes tasks, decision-making, autonomy, and authority within teams, and adjust organizational practices to prevent emerging inequalities from solidifying over time.
- Measure employee well-being as AI and internal entrepreneurial behaviors may have an impact on employees and their mental health.
