Feedback
Last date modified: 2026-May-06
Review documents
After aiR for Data Breach Response runs Data Analysis on your documents, you review the results to validate PI predictions. During review, you verify that aiR accurately identified documents containing Personal Information (PI), correct any mispredictions, and confirm that all PI values are detected. This process is similar to reviewing documents using Review Center. Reviewing documents using Review Center.
aiR for Data Breach Response categorizes documents as either Structured or Unstructured, each requiring a distinct approach to identifying PI.
Before reviewing documents, confirm that Data Analysis has completed for your document set. aiR for Data Breach Response analyzes only extracted text and does not consider metadata, images, or other non-text elements. If your documents contain image-based content, you may need to OCR them first. See OCR for instructions. All AI-generated predictions should be verified by a human reviewer.
Unstructured documents
Unstructured documents contain unlabeled or otherwise unorganized data, typically composed of natural language. Examples include:
- Emails
- Word processing documents (such as .doc and .docx files)
- Text-based files (such as .txt and .pdf files)
- Other data sources, such as photos and audio files
For unstructured documents, aiR for Data Breach Response uses the context of the document to differentiate types of PI. aiR can automatically identify and link potential PI across these documents, which may help accelerate your review and quality control processes.
To review unstructured documents, see Unstructured document review.
Structured documents
Structured documents contain data organized in a specific and predefined way, typically in a table with columns and rows where each data point has a specific data type. Examples include:
- Spreadsheets (such as .xls, .xlsx, and .xlsm files)
- Databases
- CSV files
For structured documents, aiR for Data Breach Response identifies table boundaries and analyzes header and column content to predict PI.
To review Structured documents, see Structured document review.
Frequently asked questions
What should I do if a record is marked invalid?
During Data Analysis some records may be marked invalid if they do not meet the criteria for record creation. This feature is designed to exclude “junk” records that do not contain valid entries. Invalid records will not be used in normalization and will not appear in the final Entity List. If you have a record that is invalid that you do want included in your Entity List, you should fix the record based on the Invalid Reason being generated. See Invalid Reasons for a comprehensive list of Invalid Reasons and the recommended resolution path.
What is the difference between reviewing unstructured and structured documents?
Unstructured documents rely on natural language context for PI detection, while structured documents use table boundaries and column headers. Each type has a distinct review workflow. See the relevant review topic for your document type.
Do I need to review every document manually?
aiR for Data Breach Response uses AI to predict PI across your document set. However, all AI-generated outputs should be reviewed by a human to verify accuracy. Focus your review on validating and correcting AI predictions rather than identifying PI from scratch.
