Glossary term
Glossary term
Governance and Compliance
Pattern where humans approve, review, or intervene in AI workflows.
UiPath's AI-assisted document processing routes low-confidence extractions (confidence <80%) to a human reviewer queue - an RPA bot processes 90% of invoices automatically while humans handle edge cases.
Salesforce Agentforce uses configurable HITL thresholds - refund requests above $500 are automatically escalated to a human agent with the AI's recommended resolution pre-populated for one-click approval.
The EU AI Act requires HITL for high-risk AI decisions (hiring, credit, law enforcement) - a recruitment platform uses a mandatory human review step before any AI shortlisting decision is communicated to a candidate.