Review Hub

The Review Hub is your central workspace for verifying OCR-extracted data. After documents are uploaded and processed, they appear here for human review. This human-in-the-loop workflow ensures accuracy by letting you confirm or correct extracted values before export.

Review Hub overview

The Review Hub displays all processed documents organized by client. Each document shows its form type, processing status, confidence score, and any validation warnings. You can view documents grouped by client or as a flat list.

From the work queue, you can:

• Filter by status (Ready for Review, In Review, Processing)
• Search by client name
• Sort by upload date, confidence score, or client name
• Toggle between grouped and flat list views
• Start a review session for all pending documents belonging to a client

Agent proposal review

When an enabled agent submits a suggestion, Review Hub shows it in the Agent Proposals queue as a pending, untrusted suggestion. The page keeps three value planes separate:

• Agent suggestion is the text or typed value proposed by the agent.
• Current provisional is the value currently being reviewed in the packet.
• Current approved is the last approved value, when one exists.

The suggestion never silently replaces either current value. Review the packet, source references, and any bounded evidence before choosing an outcome. A proposal may be marked Stale if the packet revision, field, form, client, or tax-year context changed after the agent created it. Refresh the proposal and re-check the current value; a stale proposal cannot be applied.

Apply accepts a field-value proposal through the normal human correction and revision workflow. It may change the provisional value and records the human outcome; it does not mean the packet is approved. Applying an observation records the observation without changing tax data. Reject leaves tax data unchanged and requires a reason category; an optional note can add context. Rejected and stale proposals cannot be applied, and terminal proposals cannot be replayed.

If an approved packet is reopened after a correction, the approved snapshot is preserved while the new correction becomes the current provisional state. Any new approval still follows the existing Review Hub approval controls.

Evidence shown for a proposal is limited to an authorized, tightly cropped, masked image region. Review Hub does not expose a full page, document download, raw OCR, unrestricted text, storage path, signed URL, or arbitrary source link through proposal review. Agent-provided text is untrusted plain text and is not rendered as HTML or executable links.

Proposal links are authenticated navigation hints, not access passes. Opening a copied deep link still requires a signed-in Review Hub session, active firm membership, and reviewer authorization for the packet. If any of those checks fail, the page returns the safe not-found/authorization state without revealing whether another firm's proposal exists.

When you click Review on a document, the split-screen review interface opens. The left side shows the original document, and the right side displays extracted fields for verification.

Review queue states

Documents move through several states as they progress from upload to approval. Understanding these states helps you prioritize your review work.

Processing states

Status	Description
Pending	Document is queued for OCR processing. Not yet ready for review.
Processing	OCR extraction is in progress. You cannot review until this completes.
Completed	OCR finished successfully. Document is ready for review.
Failed	Processing encountered an error. Check the error message or contact support.

Review states

Status	Badge Color	Description
Ready for Review	Red	OCR completed. Document needs your verification before export.
In Review	Blue	Someone has opened this document for review but has not yet approved it.
Approved	Green	Document has been verified and is ready for export.
Rejected	Gray	Document was rejected and sent back for reprocessing.

Confidence indicators

Each document displays a confidence score that reflects how certain the OCR engine was about its extractions.

Confidence Level	Score Range	What It Means
High	90% and above	OCR is very confident. Fields are likely correct but should still be verified.
Medium	20% to 89%	OCR has moderate confidence. Review these fields carefully.
Low	Below 20%	OCR had difficulty extracting these values. Manual verification is essential.

Documents with high confidence scores across all fields may be pre-approved automatically. Even pre-approved documents appear in your queue for final confirmation.

Additional status badges

You may also see these badges on documents:

• Manual: The form type was manually overridden from the original classification
• Unknown - Auto-Approved: The system could not identify the form type but approved it automatically
• Identity Incomplete: Taxpayer identity information is missing or partial

Review+ availability

Review+ delegated review is coming soon and is not currently available. Continue to assign review work to members of your own firm in Review Hub.

Field validation and edits

When you open a document for review, you see the split-screen interface with the original document on the left and extracted fields on the right.

Viewing the document

The left panel displays your original uploaded document. For PDFs, you can navigate between pages, zoom in and out, and rotate the view. For images, the viewer adjusts to fit the panel.

Use the document viewer to cross-reference extracted values against the source material.

Editing fields

Each extracted field shows:

• The field name (for example, Employee SSN, Wages)
• The extracted value in an editable text box
• A confidence badge indicating OCR certainty

To edit a field:

1. Click in the text box for the field you want to change
2. Type the corrected value
3. Click Save Changes to persist your edits

Your changes are saved immediately but do not change the document status until you approve it.

Confidence-based review

Fields display confidence badges that help you focus your attention:

• Green badges indicate high confidence (90%+). These fields are likely correct.
• Yellow badges indicate medium confidence. Review these carefully.
• Red badges indicate low confidence. These fields need your attention.

When you see a field with a low confidence score, compare the extracted value against the document viewer to verify accuracy.

Business rule violations

Some documents display rule violation counts in the queue. These indicate that extracted values failed validation checks, such as:

• SSN format errors
• Negative wage amounts
• Missing required fields
• Values that fall outside expected ranges

Rule violations appear as warnings in the field editor. Address these before approving the document.

Exception handling

Processing failures

If a document fails during OCR processing, you see a Failed status badge and an error message. Common causes include:

• Corrupted or password-protected PDF
• Image quality too low for OCR
• Unsupported file format
• Temporary service interruption

For failed documents, try uploading a higher-quality scan or a different file format. If failures persist, contact support.

Edit conflicts

If another user edits the same document while you are reviewing it, you may encounter a conflict when saving. The system alerts you to the conflict and shows both versions. Resolve conflicts by choosing which values to keep, then save again.

Identity mismatches

When processing multiple documents for the same client, the system checks for identity consistency. If names or SSNs do not match across documents, you see an identity mismatch warning. Review both documents to determine which values are correct.

Manual form type override

If the OCR incorrectly identifies a form type, you can manually select the correct type from the form type dropdown. This triggers re-extraction using the correct form processor. The document displays a Manual badge to indicate the override.

Adding missing fields

Sometimes the OCR misses fields that exist on the document. Use the Add Missing Fields control to select additional fields for extraction. This is useful when:

• A field was present but not detected
• You need to capture optional fields not in the default extraction set
• The form has non-standard field placements

Related resources

• Uploading Documents: Learn how to submit documents for processing
• Exporting Data: Export approved documents to CSV, Excel, or other formats