Research

AI is changing workplaces: it leads to higher individual productivity and faster learning. We model the effects of these changes on internal labor markets. We show that for productivity shocks, firms preserve the span in the long run: a pyramid remains a pyramid. In the short run, however, firms freeze junior hiring, temporarily transforming the pyramid toward a diamond. In the transition to the new steady state, firms may oscillate between pyramids and diamonds before settling. For learning shocks, the long-run span falls, potentially permanently shifting the firm from a pyramid to a diamond. These fluctuations create inequality between junior cohorts and affect firm value when human capital is firm-specific.

AI expands firms’ capacity to measure performance, but these measures operate through middle managers who sustain relational contracts with workers. We study how these relationships shape performance measure design. A measure better aligned with headquarters can lower the manager’s payoff from formal incentives alone, making harmful relational contracts easier to sustain. Headquarters may therefore deliberately limit the use of better-aligned performance measures. Improvements in AI-enabled measurement can thus yield limited gains even under optimal design.

A content platform sequentially contracts with artists and later controls their exposure via algorithmic recommendations. We identify a time-inconsistency problem: once an early artist is locked into a royalty contract, the platform has an incentive to sign later artists at lower rates and favor them in recommendations. Anticipating this, early artists demand compensation entirely through fixed payments, driving equilibrium royalty rates to zero. When fixed payments are relatively costly, royalties become positive but unequal, distorting recommendations. The platform may mitigate these inefficiencies by designing the sequence of artists to contract with or committing to uniform royalty rates.