> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ingestly.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Human review

> Add human-in-the-loop review to validate and correct extracted data.

Human review lets you add a verification step to your pipeline where a person reviews and optionally corrects AI-extracted data before it proceeds.

## When to use human review

* **High-stakes documents:** financial, legal, or compliance documents where accuracy is critical
* **Low-confidence extractions:** only review documents where the AI is uncertain
* **Training period:** review all documents initially, then switch to low-confidence mode as you refine your schema

## Adding a review step

1. Open your pipeline in the [editor](/pipelines/editor)
2. Add a **Review action** node after your extract action
3. Connect: Extract Action → Review Action → (next step)
4. Configure the review node

### Configuration

| Field                    | Options                                                                                                                                                                                                                                                                                                                                                     |
| ------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Trigger mode**         | `Always`: review every document. `Low confidence`: review only when AI confidence is below the threshold. `Unverified fields`: review only when the upstream Extract node could not ground one or more fields in the document. `Validation failures`: review only when an upstream Validation node (in **Mark for review** mode) flagged one or more fields |
| **Confidence threshold** | A value between 0 and 1 (only for low-confidence mode). Documents with confidence below this threshold are sent for review                                                                                                                                                                                                                                  |
| **Instructions**         | Optional. Guidance shown to reviewers (for example, "Verify the total matches the sum of line items")                                                                                                                                                                                                                                                       |
| **Auto-assign reviewer** | Optional. Defaults to "Anyone can claim". Otherwise, pre-assigns the review task to the selected teammate (reassignable later)                                                                                                                                                                                                                              |

## How the review workflow works

```mermaid actions={false} theme={null}
sequenceDiagram
    participant Pipeline
    participant Review Queue
    participant Reviewer
    Pipeline->>Review Queue: Create review task
    Note over Pipeline: Run pauses
    Review Queue->>Reviewer: Reviewer claims task
    Reviewer->>Reviewer: Review and edit data
    alt Approved
        Reviewer->>Pipeline: Approve
        Note over Pipeline: Run resumes with approved data
    else Rejected
        Reviewer->>Pipeline: Reject
        Note over Pipeline: Run marked as failed
    end
```

1. The run reaches the review action and **pauses**
2. A review task appears in the [review queue](/reviews/queue)
3. A reviewer claims and reviews the extracted data
4. On **approve:** the run resumes with the (possibly corrected) data
5. On **reject:** the run is marked as failed

## Example pipeline with review

A common pattern is to extract data, review it, then send it to a callback:

**Upload Trigger → Extract Action → Review Action → Callback Output**

This ensures a human validates every extraction before results are delivered.

## Tips

<Tip>Start with trigger mode set to **Always** when first deploying a pipeline. Once you're confident in the extraction quality, switch to **Low confidence** to reduce manual review volume.</Tip>

* Write clear, specific review instructions to guide reviewers
* Use the low-confidence threshold to balance accuracy and review workload
* Review tasks are visible to all users with review permissions in your organization