Persistent PM agents
Project managers are durable roles with memory and operating principles, not disposable prompts.
AI project management
AI Project Management
A project-management system for coordinating human decisions, persistent PM agents, AI workers, requirements, plans, and execution history.
Problem
AI workers can move quickly, but complex projects need ownership, prioritization, durable decisions, requirements discipline, and a way to keep the work aligned when multiple sessions or agents are involved.
Solution
AI Project Management treats humans and agents as participants in a managed delivery system. It emphasizes explicit requirements, persistent PM identity, work queues, decision records, and traceable execution.
Requirement discipline
Work is framed as deliverables with state, priority, verification, and explicit scope.
Human approval loops
The design preserves human authority for decisions that need Dave's judgment.
Technical highlights
Architecture
- Agent and human roles modeled as durable project actors.
- Plans, requirements, decisions, and execution history linked into a managed workflow.
- Designed to integrate with persistent memory and MCP-backed tooling.
Dave's role
Creator, approver, and product owner, defining the management model and the boundaries between autonomous agent work and human judgment.
Stack and status
Stack
- AI-agent workflow design
- Requirements and plan state machines
- MAS/MCP-style persistence concepts
- PMBOK-influenced operating discipline
Private repo / active design work
Open GitHub
Repository access is available to Dave when logged in.