Docker defines maturity levels for agentic coding adoption

The Agent Baseline framework categorizes organizational adoption of AI coding into three maturity tiers. Level 1 is characterized by individual developers using coding agents with full end-to-end direction. Level 2 involves predefined workflows where agents handle specific tasks like code generation and testing within a structured process. Level 3 represents full autonomy where agents determine the next steps, select appropriate sub-agents, and manage the workflow without predefined paths. Eli Aylena of Docker noted that while few organizations have successfully reached Level 3, many are on the journey, and the appropriate level depends on the specific task.

The framework highlights that supply chain risks in agentic development are dynamic, as agents may clone third-party repositories that contain their own agent configurations, skills, and MCP server connectors. This creates a complex attack surface that static package analysis cannot detect. Panelists argue that organizations should assess their current maturity level and consolidate workflows to improve both security and usability, rather than adopting a fragmented buffet of different coding agents. They emphasize that successful adoption requires a structured approach to managing the autonomy of AI agents within the development lifecycle.