The Redact Agent finds personal data in PDFs and scans and blacks it out: the text under every black box is removed and the original stays unchanged. A saved Redact Agent keeps all settings and processes the documents you select server-side, for example before sharing under the GDPR, for a records request or for an archive.
Prepare the documents
Put the PDFs into a workspace. The agent processes the documents you select in the document overview; scans are read with OCR.
Create the Redact Agent
Open “Document AI → Redact Agent” and click “Create Redact Agent”. On “General” give the agent a clear name and choose the AI tier: Basic for simple documents up to 15 pages, Premium for scans, tables and forms up to 50 pages, Ultra for sensitive files up to 100 pages.
Choose a profile
On “Scenario” choose one of the 13 profiles, for example GDPR request, external sharing, HR file or financial data. The profile defines which categories are blacked out and what stays visible. With the profile “Custom” you choose the categories yourself.
Define exceptions and extra terms
On “Fine-tuning” enter terms that are never redacted (“Do NOT redact (Whitelist)”, for example your company name) and terms that are always redacted (“ADDITIONALLY redact (Blacklist)”, for example case numbers). Custom AI instructions are optional.
Set the output
On “Output” choose an existing or a new workspace for the redacted copies and the box colour. Pseudonymizing instead of blacking out, that is replacing data with fictitious values, needs Ultra and is not available in every profile. Click “Save changes”.
Test with a few documents
In the document overview select two or three typical PDFs and click “Actions” → “Agents & workflows” → “Redact” → “Automatically with Redact Agent”. Pick the agent and click “START REDACT AGENT”. A single document opens the preview first, where you see every marked area with its category and reason. Check the redacted copies in the target workspace. If something is missing or too much is blacked out, adjust the profile, whitelist or blacklist of the agent and test again.
Process the batch
When the test looks right, select the remaining PDFs and start the agent the same way. Several documents are processed server-side without preview, four at a time. Follow the progress in the job monitor “Processing”. You pay per processed page, depending on the AI tier.
Automate
A workspace rule that redacts every new upload automatically does not exist. For new documents, start the saved agent from your own system through the REST API (POST /documents/redact-agent-run). The technical reference on the product page describes the call.
