Boss Class 1. Fat layer of humans

What It Covers

Andrew Palmer explores generative AI's impact on white collar work through interviews with Silicon Valley investors, academics, and AI researchers. He tests AI tools on his own writing tasks, discovers the jagged frontier where AI excels and fails unpredictably, and confronts the unsettling reality that AI-generated content can rival human work in unexpected ways.

Key Questions Answered

• •The Jagged Frontier Concept: AI capabilities form an irregular mountain range rather than a straight line, excelling dramatically at certain tasks while failing surprisingly at others. Organizations must map this terrain through experimentation in their specific domains, as the only way to identify where AI helps versus hinders is testing it against expert knowledge in each field.
• •AI as Capacity Gain Not Cost Cutting: Companies treating AI as a workforce reduction tool fundamentally misread the moment. The correct approach treats efficiency gains as capacity expansion during an industrial revolution, enabling organizations to pursue new capabilities and competitive advantages rather than simply firing people. This requires organizational research and development investment, not headcount reduction.
• •Experimental Integration Method: Professor Ethan Mollick recommends using AI for every single work task for one full day to discover its boundaries. This forced experimentation reveals unexpected strengths and weaknesses specific to individual roles. If nothing about the experience creates anxiety or surprise, the user lacks sufficient ambition in testing the technology's limits.
• •Organizational Restructuring for AI: Large companies experiment with pulling senior engineers from IT departments and forming three-person teams combining subject matter experts, engineers, and strategists. These restructured units build products in one week that previously required six months and teams of twenty people, demonstrating how AI enables radical organizational form changes beyond simple automation.
• •AI Writing Performance Reality: When Palmer conducted blind tests of his satirical column against AI-generated versions, colleagues split nearly evenly in identifying the human author. Three chose the human version for being funnier, two selected the AI version for better coherence and conversational flow. This forces writers to identify which job elements require human authorship versus where AI assistance adds value.