Human-in-the-Loop

What it means

The product pauses at defined checkpoints for human judgment (approve send, merge PR, publish article, charge card), even if AI drafted the action.

Why designers should care

HITL patterns need clear diffs, one-click approve/reject, audit logs, and SLA-aware queues, especially for agents that chain multiple steps.

Example

An email agent drafts replies in “Pending review”; support leads batch-approve five messages with inline edits before anything reaches customers.

Common mistakes

• • Approval UI that hides what changed from the AI’s last version.
• • Human review added only after an incident, with no metrics on queue depth.
• • Training users to rubber-stamp because diffs are unreadable.

Related terms

• Agent: The model decides sequences of actions (search, draft, click, call APIs) within policies until the task completes or hits a human checkpoint.
• Guardrails: Guardrails include system prompts, classifiers, allow/deny lists, output validators, rate limits, and human review gates applied before or after generation.
• Hallucination: The model generates plausible-sounding content that does not match facts, retrieved sources, or user-provided data.
• Workflow: Defined stages (inputs → generate → review → export) often spanning people, systems, and multiple model calls with clear deliverables.
• Tool Use: The model emits structured calls (name + arguments) that your app executes, then feeds results back into the conversation for the next step.

Explore more

• Agentic UX framework