Ethan Mollick

791 professionals took part in the pre-registered "The Cybernetic Teammate" field experiment at P&G, co-authored by Ethan Mollick with Dell'Acqua, Lakhani and others: an individual with AI matched the performance of a two-person team without it, and the disciplinary boundary between R&D and commercial staff collapsed. That result is the through-line in most of what he writes. Mollick, associate professor at The Wharton School and author of «Co-Intelligence», keeps returning to a single claim: the bottleneck for AI is organizational, not technical.

His earlier research with «Boston Consulting Group» (758 consultants) produced the «jagged frontier»: +12.2% tasks completed, +25.1% speed, +40% quality inside the frontier, and 60-70% accuracy versus 84% for unassisted humans outside it. From there the argument widens. In «Real AI Agents and Real Work» he predicts the next model generation will beat human experts on average. In «Giving your AI a Job Interview» he argues standard benchmarks miss task-specific performance and prescribes interviewing models on your own work. In "Organizational Theory Lessons for Agentic AI" he doubts orchestrators can supervise unlimited sub-agents and proposes middle-management agents.

The organizational claim has teeth. He measures 70% of employees experimenting with AI but only 20% using it regularly, roughly 15% adoption under mandates against 40% bottom-up, and states that HR is R&D now. His sharpest test of the labs: dissolve the forward deployed engineering teams, and one will know they believe in ASI.