The Last Human criticizes multi-agent org charts

The author argues that recreating human organizational charts with AI agents often results in performing unnecessary work without improving actual productivity or accuracy. Assigning expert personas like legal advisors or finance leads does not grant language models genuine professional competence, and multi-agent debates can easily suffer from sycophancy that reinforces errors rather than correcting them. The article concludes that these artificial organizations often function more as superficial performances of expertise than as useful business tools. Open article →