Ethan Mollick

Confidence 0.80 · 4 sources · last confirmed 2026-10-11

Associate professor at The Wharton School, University of Pennsylvania. Co-director of the AI Initiative at Wharton.

Co-author on the Jagged Technological Frontier paper (Dell’Acqua et al., Org Sci 2026), one of the canonical empirical studies of generative AI’s effect on knowledge work — the source of the jagged frontier term (jagged-frontier). Author of Co-Intelligence (and the forthcoming Co-Existence) and the One Useful Thing newsletter; the wiki’s deliberately non-doomer, non-zealot practitioner voice.

In the A Bit of Optimism interview (with Simon Sinek, June 2026) Mollick lands several of the wiki’s live theses from the social-science chair: experience beats “AI-native” youth (the durable-skills / returns-to-expertise finding, via the BCG junior-worse-at-AI study); taste as the differentiator as output commoditises (“the death of the movie star”); the apprenticeship model just broke (the ai-deskilling talent-pipeline mechanism); the model/app/harness three-layer framing (agent-harness); and a heavy reliance on GDPval (2025-10-05-patwardhan-et-al-openai-gdpval, 48%→84%) for “how good the work is.”

The study behind the soundbite (added 2026-09-15)

Mollick is a co-author on [[2026-06-12-dellacqua-cybernetic-teammate-field-experiment-genai-teamwork|The Cybernetic Teammate]] (Organization Science, 2026, with Fabrizio Dell’Acqua and Karim Lakhani), which supplies the measurement behind the claim he makes for general audiences in his Simon Sinek interview — that evaluation, not generation, is the bottleneck. The paper’s decomposition finds AI “primarily enhances the quality of generated ideas… whereas human judgment retains value in evaluative selection.”

The corpus now holds Mollick at both altitudes: the peer-reviewed experiment and the popularisation of it. Useful when weighing his public claims — on this one, the underlying evidence is preregistered and N=791.

On HBR IdeaCast: strategy, not productivity (added 2026-10-11)

In HBR IdeaCast (October 2026), introduced as “author of Co-Existence”, Mollick moves from individual use to organisational design: productivity gains “get eaten up by the organization” unless KPIs change; the leadership, lab and crowd model; delegating to agents as a “conjurer”; twilight rather than dark software factories; and four capabilities for leaders (deep knowledge, wide knowledge, taste, agency). His claim that less-experienced BCG consultants were worse at using AI sits awkwardly beside his own jagged-frontier paper’s larger gains for lower performers; the source page records the tension.