AI (RAG) indexing
Uploaded documents are automatically processed for semantic search and AI question-answering, so the content — not just the filename — becomes something the AI Copilot can search and reason about.
-
When a file is uploaded (or a new version is added), it’s queued for indexing. Supported formats (PDF, DOCX, and plain text; scanned images may require OCR) have their text extracted and embedded.
-
Once indexed, the file’s content is searchable through AI Copilot and, where applicable, surfaced by the Documents Librarian agent when it looks for related or duplicate content.
-
Indexing respects sensitivity and permissions at query time — the AI never surfaces content from a file the asking user isn’t allowed to see, even if that content is semantically relevant to their question.
-
Very large files or unsupported formats may not be fully indexed; the AI Hub shows indexing status for files in your folders.