Autonomy-level classification
An autonomy-level classification describes how independently an AI system or agent can make decisions and take actions without human direction or intervention.
AI Governance
A grade describing how independently an AI system or process can operate from human direction.
- How it is used
- Sets action-specific autonomy according to potential harm, reversibility, and available oversight.
- How deep it goes
- Details tier assignment and enforcement; identifies reported autonomy Levels 1–5 without defining each higher level.
- Examples
- Uses non-sensitive data reads, production edits, payroll access, and a reported database deletion to illustrate autonomy decisions and risks.
- What you can do after
- Evaluate and configure action-specific autonomy tiers with approval gates, agent credentials, and a kill switch.
Read how AI Governance explains autonomy-level classification →
Designing AI Agents
The degree to which an agent can make decisions and take actions without human intervention or authorization.
- How it is used
- Classifies execution authority and human oversight to identify the current level, intended next level, and controls needed for the transition.
- How deep it goes
- Defines levels 0–4 as operating arrangements and level 5 as a research horizon, while distinguishing the working scale from an industry standard.
- Examples
- Illustrates levels with shadow-mode review, a read-only assistant, a coding agent requiring approval, and a bounded background agent.
- What you can do after
- Evaluate and understand an agent’s authority by naming its level, hard ceiling, task risk, and evidence for changes in trust.
Read how Designing AI Agents explains autonomy-level classification →
What to know first
- human-in-the-loopHuman-in-the-loop is a system or workflow design in which people participate in reviewing, guiding, correcting, or authorizing an automated process.
- prompt injectionPrompt injection is an attack that introduces instructions through input or context to manipulate an AI system’s behavior or override its intended instructions.
- chain-of-thoughtA chain-of-thought is an ordered sequence of intermediate reasoning steps that connects information or a problem to a conclusion, whether expressed or generated as text.
- guardrailA guardrail is a constraint, control, or monitoring mechanism that limits system behavior to reduce risks and keep actions or outputs within acceptable bounds.
Related concepts
- least agencyconfigured by · A principle of granting an agent only the autonomy needed to perform its delegated task.
- autonomyconfigures · Autonomy is the capacity or degree to make decisions and act independently within defined responsibilities and constraints.
- evidenceproduced by · Evidence is information, observations, or material used to support, challenge, or assess a claim, conclusion, event, or outcome.
Questions about autonomy-level classification
What is autonomy-level classification?
An autonomy-level classification describes how independently an AI system or agent can make decisions and take actions without human direction or intervention.
Which Manning books explain autonomy-level classification?
AI Governance and Designing AI Agents.
What should I know before learning autonomy-level classification?
Manning books that teach autonomy-level classification usually introduce human-in-the-loop, prompt injection, chain-of-thought and guardrail first.
What is autonomy-level classification part of?
Autonomy-level classification is part of autonomy taxonomy. A book-specific scale of operating arrangements distinguished by execution authority and human oversight, with levels 0–5.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.

