Framework
The Human-AI Task Scale
Seven levels that describe how a task is shared between a person and AI. Each task gets a level for today and a proposed level for tomorrow.
- L1Human OnlyNo AI
- L2AI AssistSurfaces options
- L3AI Co-pilotAI drafts, human edits
- L4Human-Led CollaborationShared iteration
- L5AI-Led with Human ReviewEvery output approved
- L6AI Autonomous with OversightExceptions only
- L7Fully Autonomous AgentHuman sets policy
Why levels, not yes or no?
"Should AI do this?" is the wrong question. Almost every task can involve AI to some degree. The useful question is how much, and who stays accountable.
The scale makes that answer specific. Drafting a memo can move to L5, where AI writes and a person approves every output, while the final approval stays at L1. Different tasks in the same role land on different levels, which is why roles change one task at a time.
The Human-AI Task Scale is owned by Talent Rewire, Inc. It first appeared in print in Applying AI in Learning & Development (ATD Press).