iManage AI Enrichment processes each document through three stages: Classify, Detect, and Extract. This page describes the Classify stage.

The iManage AI Enrichment document classification model automatically identifies and labels documents using a three-category top-level classifier: Agreement, Court Document, and Undetermined. This classification provides a consistent way to categorize documents within the workspace (these may also be called ‘matters’ or another name, depending on your configuration and application).

The top-level categories are defined as follows:

  • Agreement: A legally binding document or arrangement, whether unilateral, bilateral, or multilateral, that sets out terms, conditions, rights, or obligations intended to have legal effect within a specific context or situation.

  • Court Document: Written materials and records associated with legal proceedings, used to document cases, inform parties of legal actions, and support the administration of justice.

  • Undetermined: Documents that don't fall within the Agreement or Court Document classifications. This includes legal and non-legal materials whose nature, type, or purpose doesn't align with a supported agreement type or court-related record.

NOTE: Classification is based on a document’s actual content, not its file type or origin, with one exception: an email saved as a PDF is always classified as Email, regardless of what the email itself discusses. For more information, refer to Why are some emails classified while others aren’t in the FAQ.

For documents classified as Agreement, iManage AI Enrichment then runs a Detect stage. Detect identifies the agreement’s specific sub-type (there are 66 supported sub-classes for Agreements). Detect runs automatically and doesn’t produce a separate visible property.

After Detect completes, iManage AI Enrichment applies additional AI-enriched metadata properties. These properties capture key agreement-specific information (such as Agreement Date, Jurisdiction, and Heading) and enable more detailed understanding, searching, and filtering of agreement documents.

The enrichment model is trained on documents from multiple jurisdictions and is applicable across a wide range of legal practice areas.

For documents categorized as Court Document or Undetermined, only the Category property and the administrative properties are set; these documents don’t proceed to the Detect or Extract stages. No agreement-specific enriched metadata properties are populated for these categories.