# Human-in-the-Loop (HITL)

> Human-in-the-loop is a design where AI handles the routine work automatically but hands off to a person for the cases that need judgment, approval, or oversight — combining speed with accountability.

_Source: https://plenaura.com/glossary/human-in-the-loop · Last updated: 2026-06-03 · Plenaura_

## Key points

- AI does the bulk of the volume fast; people focus only on the uncertain, high-stakes, or unusual cases.
- Confidence scoring decides what to auto-process and what to escalate, so nothing risky slips through unchecked.
- Every human decision is a training signal — the system gets more accurate and escalates less over time.
- It's how Plenaura builds AI that's trustworthy in production: automate the routine, keep a person on the judgment calls.

## FAQ

### Doesn't keeping a human involved defeat the point of automation?

No — the goal is to remove routine effort, not all oversight. The AI handles the high-volume, low-risk work on its own and only involves a person for the small share of cases that genuinely need judgment, which is where human time is most valuable.

### How does the system decide when to involve a human?

It attaches a confidence score to each decision. Anything above your threshold is processed automatically; anything below it, or in a defined high-risk category, is routed to a person to review. You set where those lines are based on your risk tolerance.