Skip to main content

Autonomous Operation

1. Definition​

Autonomous is an execution mode in which the platform can continue toward an objective with limited human intervention, repeatedly assessing state, planning, acting, observing, and re-planning within explicit boundaries.

It should not be defined as "an agent that runs automatically."

flowchart TD
    O(["Objective"]):::actor --> S["Assess State"]:::process
    S --> P["Plan"]:::model
    P --> G{"Policy Check"}:::govern
    G -->|"Allowed"| A["Act"]:::process
    G -->|"Approval required"| H["Human Approval"]:::govern
    G -->|"Denied"| X(["Stop / Escalate"]):::govern
    H -->|"Approved"| A
    H -->|"Rejected"| X
    A --> V["Observe"]:::process
    V --> D{"Outcome"}:::gate
    D -->|"Objective met"| C(["Complete"]):::output
    D -->|"Exception"| X
    D ==>|"Continue"| S

    classDef actor fill:#DBEAFE,stroke:#1D4ED8,stroke-width:1px,color:#0B1220
    classDef process fill:#EEF2F7,stroke:#475569,stroke-width:1px,color:#0B1220
    classDef model fill:#EDE9FE,stroke:#6D28D9,stroke-width:1px,color:#0B1220
    classDef gate fill:#FEF3C7,stroke:#B45309,stroke-width:1px,color:#0B1220
    classDef govern fill:#FFE4E6,stroke:#BE123C,stroke-width:1px,color:#0B1220
    classDef output fill:#DCFCE7,stroke:#15803D,stroke-width:1px,color:#0B1220

2. Autonomy Is a Governance Problem​

Autonomy increases the need for:

  • Explicit permissions.
  • Action boundaries.
  • Approval policy.
  • Stop conditions.
  • Step/time/cost limits.
  • State persistence.
  • Failure recovery.
  • External side-effect control.
  • Audit and traceability.
  • Escalation paths.
LevelBehaviorExample
A0No autonomyAssistant answers only
A1SuggestSystem recommends actions, user executes
A2Approve-to-actSystem plans; user approves side effects
A3Bounded autonomousSystem executes allowlisted actions and escalates exceptions
A4Extended autonomousSystem manages a continuing objective across multiple systems within policy

This level model is a recommended Site governance mechanism, not an external standard.

4. Autonomous Composition​

An autonomous capability normally composes other capabilities.

flowchart TD
    C[["Autonomous Controller"]]:::model --> A[["Agent: analyze and decide"]]:::model
    C --> W[["Automation: execute repeatable process"]]:::process
    C --> S[["Assistant: communicate with humans"]]:::model
    C --> P["Rules / Policy: approval boundary"]:::govern
    C --> E["Scheduler / Event loop: continue operation"]:::process
    C --> ST[("State store: objective and progress")]:::store

    classDef process fill:#EEF2F7,stroke:#475569,stroke-width:1px,color:#0B1220
    classDef model fill:#EDE9FE,stroke:#6D28D9,stroke-width:1px,color:#0B1220
    classDef store fill:#CCFBF1,stroke:#0F766E,stroke-width:1px,color:#0B1220
    classDef govern fill:#FFE4E6,stroke:#BE123C,stroke-width:1px,color:#0B1220

5. Example​

Objective: Keep deployed Site application packages compliant with approved versions.

Possible loop:

  1. Monitor package/version events.
  2. Compare deployed and approved versions.
  3. Agent assesses whether change is required.
  4. Retrieve impact and dependency information.
  5. Policy determines whether approval is required.
  6. Approved automation updates the package.
  7. Validation verifies the deployment.
  8. Exceptions are escalated.
  9. State is persisted and monitoring continues.

6. Required Stop Conditions​

Every autonomous capability must define at least one bounded completion or stop mechanism:

  • Objective achieved.
  • Maximum iteration count.
  • Maximum elapsed time.
  • Cost/token budget reached.
  • Unrecoverable error.
  • Permission denied.
  • Human stop command.
  • Policy violation.
  • External system unavailable beyond retry threshold.

7. What Is Not Autonomous​

  • A cron job.
  • A fixed n8n workflow.
  • A chatbot launched automatically.
  • A one-shot agent run without continuation.
  • A workflow with no dynamic planning.

Autonomy requires continued objective-driven control, not merely background execution.