The New AI Org Chart

What It Covers

Jack Dorsey and Sequoia's Roelof Botha publish an essay arguing Block is replacing hierarchical management with an AI "world model" that routes information instead of humans. The AI Daily Brief pairs this top-down architectural vision with Every's bottom-up lived experience of running a half-agent company.

Key Questions Answered

• •Hierarchy as information protocol: Every organizational layer from Rome's 8-soldier contubernium to modern middle management exists solely to route information up and down chains of command — not to add value directly. AI can now perform this routing function, making the structural rationale for traditional management layers obsolete. Recognize which managers primarily relay information versus generate decisions.
• •Block's three-role model: Dorsey collapses all job functions into three categories: Individual Contributors who build capabilities, Directly Responsible Individuals who own specific outcomes for defined periods (e.g., 90-day merchant churn reduction), and Player Coaches who develop people while still doing hands-on work. Eliminating permanent middle management layers is the explicit structural goal.
• •Emergent parallel org charts: At Every, personal AI agents naturally specialize to mirror their human owners without deliberate design. Growth lead Austin's agent becomes the de facto growth resource; product builder Dan's agent handles bug reports. Organizations should map which agents hold which institutional knowledge before that shadow org chart becomes unmanageable.
• •Personal ownership as trust infrastructure: When an individual owns a specific agent, their professional reputation backs every output that agent produces in shared channels. This reputational accountability creates organizational trust that centralized, generic AI tools cannot replicate. Assigning named agents to named humans — rather than deploying shared tools — builds credibility into AI-generated work.
• •Capability gap is imagination, not technology: Every's COO had voice-calling and email access configured in his agent for weeks before a commute forced him to actually use it. The adoption barrier is a limiting belief about what delegation is possible, not missing features. Teams should run structured "what if my agent did X" exercises to surface unused capabilities already available in deployed tools.