Five Levels of AI Autonomy
Adapted from the SAE J3016 standard for vehicle automation, AI autonomy can be categorized into five levels:
| Level | Name | Description |
|---|---|---|
| 0 | Manual | No automation; humans do everything |
| 1 | AI-Assisted | AI provides decision support; humans decide and act |
| 2 | Semi-Autonomous | AI acts with human approval for significant actions |
| 3 | Conditionally Autonomous | AI executes defined tasks; humans intervene for exceptions |
| 4 | Fully Autonomous | AI manages end-to-end; humans govern and audit |
The progression is not merely technical, it’s about trust transfer. Each level represents a different allocation of cognitive load and accountability between human and machine.
Most effective deployments don’t pick a single level. They assign different levels to different tasks based on risk, complexity, and the AI system’s demonstrated reliability.
Related: 05-atom—hitl-autonomy-inverse, 05-atom—trust-function-definition