Auto-Classification

Auto-classification is the practice of letting software, rather than a person, decide where a record belongs in an organization’s classification scheme or file plan. Using rules, pattern matching, or machine-learning models trained on representative samples, the system reads a document’s content and context and assigns it a category, which in turn drives retention, disposition, and security handling. It matters because the volume of electronic records far outstrips the capacity of staff to file each one by hand; consistent, scalable classification is what makes defensible retention and timely disposition possible at all. For example, an email mentioning a contract number and a deadline might be routed automatically to a procurement category with a seven-year retention period. Auto-classification differs from simple full-text search because its goal is durable, governed categorization tied to a schedule, not one-time discovery. Programs should validate accuracy against a human-reviewed baseline and document the methodology, since NARA’s modern guidance—reflected in the Universal ERM Requirements and FERMI after it revoked its DoD 5015.2 endorsement in 2022—emphasizes outcomes and auditability over a single prescribed tool.