The Eight Levels of AI Adoption

Mike Taylor maps AI adoption across eight levels, beginning with chatbots and progressing through copilots, agents, autopilot, workflows, proactive assistants, multi-agent systems, and orchestrators.¹

Each level represents more work delegated to AI and greater trust in its ability to operate without intervention. Taylor says higher levels are not necessarily better, with the appropriate level depending on the task, risk, and required supervision.¹

Taylor developed the framework partly by modifying Steve Yegge’s Gas Town model, an open-source system capable of coordinating dozens of coding agents with persistent memory, parallel execution, and session handoffs.² Similar orchestrator-worker architectures are already in production: Anthropic’s Research system uses a lead agent to create parallel subagents for different parts of a research task.³