This table provides a quick reference to the five levels of the Agentic AI Progression Framework as described by Pascal Bornet et al in the book Agentic Artificial Intelligence. It condenses each level into a short phrase and a brief description, moving from basic automation to full autonomy.
| Level 1 | Level 2 | Level 3 | Level 4 | Level 5 |
|---|---|---|---|---|
| Rule-based automation | Intelligent automation | Agentic workflows | Semi-autonomous agents | Fully autonomous agents |
| Systems follow predefined rules and execute only when explicitly triggered by a human. | At this level systems use basic AI to handle more complex scenarios, enabling them to process semi-structured data, recognize patterns, and make basic decisions. | AI coordinates and executes multi-step processes within a defined scope, acting without direct prompts when conditions are met. | Agents operate across a broad range of tasks, adapting to changing conditions with minimal human oversight. | AI operates entirely independently, setting its own goals and determining how to achieve them across multiple domains. |
This high-level view offers a simple way to place any AI system on the spectrum of agency. While each level brings new capabilities, it also introduces new considerations for governance, ethics, and human oversight.