Documentation Index

Fetch the complete documentation index at: https://docs.aifabrix.ai/llms.txt

Use this file to discover all available pages before exploring further.

Document approval and governance

Prev Next

When enterprises connect file libraries and content systems, not every document is ready for AI on first sync. Classification may be missing, a contract may not yet be linked to the right customer or deal, or enrichment confidence may be too low for automated use. Document approval is the governed human step that completes metadata before a document enters search, retrieval, and assistant context.

Why it matters

Unreviewed documents create compliance and quality risk. Assistants and search should not treat a file as business-ready when ownership, sensitivity, or required relationships are unknown. Document approval turns ingestion into an accountable workflow:

Document arrives from external library
  ↓
Platform syncs metadata and optional enrichment
  ↓
Required fields or links are missing or low-confidence
  ↓
Document enters Needs Review queue
  ↓
Authorized approver completes governed metadata
  ↓
Document becomes eligible for search and AI use

This is not a generic ticketing system. Approval is bound to the document-storage contract — the same business metadata, dimensions, and foreign keys that Operational Trust enforces elsewhere.

What approvers do

Approvers with document read access work from the Document Approval area in the product UI. They see a queue of documents waiting for review. When the queue is empty, that is a success state — no hidden backlog.

For each pending document, the review experience shows:

  • Original metadata from the source system (read-only context)
  • Recommended values from sync and optional enrichment
  • Required governed fields declared in integration configuration (dimensions, classifications, links to records)
  • Validation blockers when approval cannot proceed yet (for example an unresolved required link to a deal or customer)

Approvers save corrections as draft metadata, resolve foreign-key links through governed search, and approve only when eligibility rules pass. Low-confidence enrichment never bypasses human review when the integration requires it.

How governance connects

Document approval sits between Enterprise Knowledge and Operational Trust:

Layer Role in approval
Enterprise Knowledge Defines what the document means — resource type, dimensions, links to deals or customers
Operational Trust Defines who may see, edit, and approve — roles, ABAC, protection
Document storage sync Brings files and metadata in; may flag Needs Review
Approval workflow Commits canonical metadata before embed/search exposure

Required links are manifest-driven, not hardcoded per vendor. If your integration declares that every policy document must link to a customer record, the approval UI enforces that link before Approve is enabled. The platform does not invent provider-specific pickers — behavior comes from declared foreign keys and dimensions.

Sync status in business terms

Operators and architects should align on these states:

Status Business meaning
Pending / processing Document is ingesting; not ready for review
Needs Review Human must complete or confirm metadata
Validated Approver committed governed metadata
Synced Ready for governed search and assistant retrieval
Failed Ingestion error — operations or integrator follow-up

Documents can reach Needs Review because required metadata is absent, a required relationship is unresolved, or enrichment confidence is below the configured threshold. The approval queue is the control point — not an afterthought.

Audit and accountability

Every save and approval should be traceable for compliance: who changed which governed fields, when approval occurred, and whether validation blocked an action. Audit complements dimensional access control — users who cannot see a document never appear in its approval history.

Email or in-product notifications may alert approvers when new documents arrive. An empty queue plus working notifications means the process is healthy, not unused.

Who does what

Responsibility Business owner Data steward / approver Admin Integrator
Define required metadata and links Yes Advises Supports Implements in manifest
Assign approver roles Approves Yes Supports RBAC
Review and approve documents Sets policy Yes Sometimes No
Configure sync and processing flags No No Sometimes Yes
Certify corpus for AI exposure Approves Advises Coordinates Validates

Operator steps: Operate document approval. Integrator manifest and validation: Configure document approval.

Relationship to other contracts

Document storage handles files and metadata. Record storage handles structured entities. Approval often bridges them — linking a contract PDF to the correct deal or customer record before assistants may cite it.

Vector storage adds semantic search on top of an approved corpus. Do not certify RAG exposure until approval rules and dimensions match your governance pack.

Limits

Approval UX and notification channels evolve with platform releases. Required field names and dimension catalogs are integration-specific — this article describes the model, not your tenant's manifest.

Bulk metadata operations follow the same atomic and audit rules as single-document save when enabled; confirm current platform behavior before automating mass updates.