Human-on-the-Loop (HOTL)

What is HOTL?

Human-on-the-Loop (HOTL) is a governance pattern for continuous monitoring. Agents run autonomously while humans observe and can intervene if issues arise.

When to Use HOTL

• Production Monitoring - Watch live agent behavior
• Anomaly Response - React to unusual patterns
• Gradual Rollout - Monitor new agent versions
• High-Volume Systems - Where HITL isn't practical
• Learning Phase - Build trust in new agents

Enabling HOTL

Via SDK

from turingpulse_sdk import instrument, GovernanceDirective

@instrument(
    name="trading-agent",
    governance=GovernanceDirective(
        hotl=True,
        escalation_channels=["slack://trading-alerts"],
        severity="high",
    )
)
def execute_trade(symbol: str, quantity: int):
    return trading_service.execute(symbol, quantity)

Via UI

1. Navigate to Governance → Policies
2. Click Create Policy
3. Select Human-on-the-Loop (HOTL)
4. Configure alert channels and thresholds
5. Save and enable

HOTL Alerts

HOTL generates alerts when:

• Anomalies Detected - Unusual behavior patterns
• Error Spikes - Sudden increase in failures
• Threshold Breaches - KPIs exceed limits
• Drift Events - Behavior changes detected

Intervention Options

HOTL Configuration

Combining with Controls

HOTL works best when combined with:

• KPI Thresholds - Define what triggers alerts
• Drift Detection - Monitor for changes
• Anomaly Rules - Detect outliers
• Alert Channels - Route notifications

Next Steps

• Human-in-the-Loop (HITL)
• Human-after-the-Loop (HATL)
• Governance Overview