Autonomy Level

Autonomy level is the degree to which an AI employee acts without human approval — from 'drafts only' to 'fully autonomous on irreversible actions' — usually configured per task type.

Autonomy is a dial, not a switch. A common scale: L1 drafts only (every action approved), L2 sends low-stakes (high-stakes approved), L3 sends most (only spend/legal gated), L4 fully autonomous within bounded domains. Most production deployments mix levels: L1 on customer comms, L4 on internal data tasks.

Example

AI SDR runs L4 on internal CRM updates, L3 on follow-up emails, L1 on first-touch outreach. Same employee, different gates.

How OpenLabor uses it

Every OpenLabor mission has an autonomy level you set independently. Easy to start strict and relax over time.

What level should I start at?

L1 (drafts only) for the first 1–2 weeks. Promotion comes from observed quality, not promised quality.

Related: approval-gate, guardrails, ai-employee.

AI Labor Glossary