A model for human governance of AI systems.
Human-in-the-Loop is no longer one thing.
As AI systems move from drafting outputs to taking actions, organisations need clearer language for review, monitoring and accountability.
The problem
"Human-in-the-Loop" is one of the most repeated phrases in AI governance.
The problem is that people often use it to mean different things.
Sometimes it means a human reviews every output.
Sometimes it means a human monitors the system.
Sometimes it means a human designed the process and remains accountable for the outcome.
Those are not the same governance model.
Loop names three patterns: Human-in-the-Loop (review each output), Human-on-the-Loop (monitor the system), and Human-Accountable-for-the-Loop, HAL (own the system and its outcomes). Each asks a different question and fits a different class of workflow.
The three models
Review, monitor, own
Three distinct patterns for how humans govern AI-enabled workflows.
Human-in-the-Loop
AI produces an output. A human reviews it. Action follows.
AI → Human Review → Action
Human-on-the-Loop
AI operates within defined boundaries. Humans monitor and intervene.
AI → Action ↓ Human Monitoring
HAL
Humans define authority, risk, escalation and accountability. The system acts within those controls.
Owner → Policy → AI Workflow → Action → Evidence → Review
The key distinction
- Human-in-the-Loop asks:
- Did a human review the decision?
- Human-on-the-Loop asks:
- Is a human monitoring the system?
- Human-Accountable-for-the-Loop asks:
- Who owns the system that made the decision?
When each model fits
Match governance to the workflow
| Workflow type | Recommended model |
|---|---|
| Drafting a memo | Human-in-the-Loop |
| Summarising a contract | Human-in-the-Loop |
| Monitoring regulatory updates | Human-on-the-Loop |
| Screening incoming matters | Human-on-the-Loop or HAL |
| Drafting a client communication | Human-in-the-Loop |
| Routing client requests | HAL |
| Updating matter records | HAL |
| Triggering external communications | HAL with approval gates or strict controls |
| Filing regulatory documents | HAL only with high score and approval gates |
When review does not scale
HAL is the framework for the third model
Human-Accountable-for-the-Loop requires more than a named owner. HAL provides the detailed guidance, scoring system, worked examples and calculator for workflows where individual review does not scale.
Accountability cannot be delegated. Execution can.
Eight domains of accountability: ownership, authority, limits, escalation, evidence, monitoring, review, and liability.
HAL ScoreAssess whether a workflow has sufficient accountability maturity to delegate actions to AI.
Workflow CalculatorIdentify the level of governance required before you build or deploy.
Which model fits your workflow?
Start with the Loop model to identify the right pattern. If your workflow involves action, autonomy, scale or accountability risk, take the HAL assessment.