AI Document Management Agent

Description

Twenty-two years of project documentation across a file share, an old SharePoint and two ex-employee OneDrives. Nobody has audited it. We know there is duplication and we suspect there is material in there that should have been deleted under our retention policy years ago. The first job is classification: what is this document, which project does it belong to, does it contain personal data, and what is its retention class. The second job is a review queue — we are not letting an agent delete anything, but we do want it to tell us what it thinks should go, with a reason. Scanned documents are a real part of this. A meaningful share of the older material is images.

• Ingest and OCR the scanned material, reporting what could not be read rather than skipping it silently
• Classify by document type, project and retention class, with a confidence per field
• Detect personal data and near-duplicate copies across the four sources
• Build a review queue where a human approves or rejects every proposed deletion
• Produce a retention exposure report for our data protection officer

• Classification pipeline covering all four sources, including OCR
• Review queue with approve/reject and a full audit trail
• Near-duplicate and personal-data reports
• Retention exposure summary for the DPO