Automatic PII detection

Automatic detection finds personal data — names, addresses, IBANs, phone numbers, email addresses, dates, tax numbers and much more — and, if you wish, adds them all at once to the dictionary. You can start it at any time through Detect PII (the Analysis ribbon group); for a new case the assistant opens automatically after import.

Offline

Every protection level works offline on this computer — no data is transmitted to the internet during detection.

Step 1: choose a protection level

The protection level determines which categories are redacted — the search itself always runs at full depth. The level does not switch detectors off but filters the result; each level additionally redacts everything that the previous one redacts. These are the same levels as in the web interface of Docuflair Redact & Mask.

Protection level	What is captured
Minimal	Special categories only (health data or religion, for example) plus identity and access credentials. Minimal intervention — the document stays largely readable.
Standard (recommended)	Additionally all direct identifiers: names, addresses, dates of birth, contact details, bank details. The normal case for internal sharing and for sending documents out.
High	Additionally quasi-identifiers and business identifiers: places, postcodes, dates, file references, companies, registration numbers. For publication, court files and sharing with external AI.
Custom	No level filtering — only the selection of search patterns counts (the groups People, Address/Geography, Communication, Finance and others). The only way to redact monetary amounts.

If the case has custom detections, they appear as a separate section under every protection level; the “Apply this case’s custom detections” checkbox switches them on or off for that run. If the case has none yet, create them directly through Create detections….

Detection options

The gear icon at the top right takes you to the detection options:

• Language of the document — the default is to detect it automatically from the document.
• Case context:
  • Country selects country-specific classifications (US, CH, UK and so on) and the name base used for detection.
  • Subject areas (medicine, law, finance and others) unlock domain-specific detections. If detections shipped with the application are available for a newly selected subject area, the dialog offers to adopt them into the case as custom detections.
  • Redact dates only as dates of birth — for documents in which dates are needed for the subject matter (a medical record is a timeline).
  • Protection levels under which custom detections apply — the preset for the checkbox in the assistant, per level.
• Post-processing: seven post-processing steps, such as signature detection, name propagation or the person validator.

These options are stored with the case and apply to every further detection run on this document. Without an open case they only apply until the document is closed.

Step 2: review the matches and adopt them

Next starts detection; afterwards you review the matches:

1. Review the matches. Each row shows the category (coloured badge), the value found, the page, the level (sensitivity) and the proposed action. Entries that already exist in the dictionary are marked and deselected by default.
2. Make your selection. The checkbox deselects matches that should not be adopted; Select all and Select none work on the whole list. By keyboard it is quicker: the space bar toggles the checkbox of the highlighted row and jumps to the next one. If several rows are highlighted (Shift or Ctrl click), the space bar toggles them together.
3. Set the action. Per match, Redact or Pseudonymise:
   • All: redact sets every match to redaction.
   • Names & addresses: pseudonymise sets personal and company names, addresses, postcodes and places to pseudonymisation. Other matches such as a date of birth, an IBAN or a phone number cannot be pseudonymised and are redacted.
   • Names and places are set to pseudonymisation by default with a substitute word already proposed (the “Substitute word (alias)” column, editable). For redaction matches you can enter a substitute word manually to pseudonymise instead. Selected pseudonym matches without a substitute word prevent completion.
4. Finish. The selected matches are adopted as entries into the case’s target dictionary — the status bar reports, for example, “27 PII matches adopted into the dictionary.”, and the document shows the result immediately.

Whether matches are proposed for redaction by default or (for names, companies and addresses) for pseudonymisation is set under Settings → Redaction. Without the Mask licence option, only redaction is available.

Tip

Automatic detection does not replace human review. Check the result page by page — especially with scanned documents, tables and handwriting — and add anything missed with the manual tools.