Wharton

Wharton, the business school of the University of Pennsylvania in Philadelphia, enters the argument over AI adoption through one professor, Ethan Mollick, and through a complex business simulation the school built.

In May 2025 Mollick named a paradox: individual workers report large time savings, some claiming AI tripled their productivity, while their organizations record minimal gains. Between 30 and 40 percent of American workers use AI professionally; official corporate systems show engagement closer to 20 percent. The gap is shadow usage. His answer split the work between Leadership, which sets a vision and incentives for open adoption, a Lab that turns discovered workflows into scalable solutions, and the Crowd of experienced workers finding applications by trial and error.

By February 2026 he had turned to agentic systems, arguing that organizational theory applies directly. A human manager caps out below ten effective direct reports; Mollick estimates a hundred sub-agents far exceeds an orchestrator's capacity, and proposes intermediate "agent middle management". He also imports boundary objects, coupling, and bounded rationality, and doubts the AI labs see this dimension.

In May 2026 he pressed those labs: we will know they believe in ASI when they dissolve their forward deployed engineering teams. roon, an OpenAI employee, answered that the obstacle is Hayekian, information being tacit and distributed. Mollick conceded the point, then turned it back: if AI does not adopt itself, the labs' own prediction that most white-collar jobs go by 2035 is refuted by their own hiring. The adoption bottleneck migrates from the technical to the organizational.