Designing an AI company's organizational structure in Paperclip
Paperclip models a company as a hierarchy of AI agents with roles, responsibilities, and reporting lines. A proper organizational structure ensures that tasks are decomposed correctly, avoid duplication, and are performed by the right agent.
Design principles
Roles and specialization: Agents are specialized—a broad, generalist agent is inferior to several specialized ones. Researcher is separate from Writer, which is separate from Editor. Specialization improves quality through more precise system prompts and tools.
Hierarchy: The top level is Manager/Coordinator agents with access to scheduling and delegation. The bottom level is Executor agents with specific tools and limited scope. Depth: 2–3 levels is optimal.
Powers and budgets: Each agent has a limit: max tokens per task, max funds for the external API, and max actions without confirmation. If these limits are exceeded, they are escalated up the hierarchy.
Typical organizational structures
Content Marketing Team: CMO Agent → [Research Agent, Writer Agent (x3), SEO Agent, Publisher Agent]
Development Team: CTO Agent → [Architect Agent, Backend Agent (Claude Code/Codex), Frontend Agent (Cursor), QA Agent, DevOps Agent]
Support Team: Support Lead → [Triage Agent, L1 Support (x5), L2 Support (x2), CRM Agent]
Documentation
Deliverable design: organizational structure diagram, agent cards (role, tools, authority, budget), escalation matrix, runbook for typical tasks.