Back to the product map

Knowledge and control

Turn approved material into reusable knowledge without flattening access

Author working documents, organize the sources the organization relies on, retrieve answers with visible citations and keep sensitive actions proportional to each person's role. Audit and recovery make change inspectable without presenting automation as authority.

Operating outcome

Reuse what the organization knows while preserving who can see, change and recover it.

Documents, the knowledge base, cited knowledge chat, roles, activity history and recovery points solve different parts of control. Their separation is deliberate: authoring is not indexing, a citation is not an approval and a recovery point is not permission to restore.

Operating outcomes

What this product area helps make visible

These outcomes connect the implemented capability to a practical operating change. They are review paths, not guaranteed performance claims.

Human journey

From authored material to a governed organizational memory

Write the material, organize access, retrieve the approved source when work needs it and preserve evidence or recovery options for consequential change.

  1. Author the working material

    Create a personal or organization document in the native workspace and preserve revisions as the content changes.

  2. Organize scope and access

    Place documents in visible folders, share them at an explicit access level and group indexed sources into collections that people can recognize.

  3. Retrieve with the source attached

    Search the organization's indexed passages directly or ask the knowledge workspace for an answer that names the source documents it used.

  4. Inspect and recover consequential change

    Use role permissions and activity history to understand authority, then preview and confirm recovery work through its separate protected workflow when needed.

Distinct controls for authoring, retrieval, access and recovery

The platform implements a scoped documents workspace, an indexed knowledge base, source-cited knowledge answers, role-aware activity history and organization recovery points. Each surface keeps its own permission and feature boundary.

Personal and organization documents with folders, sharing and versions

Draft structured documents in personal or organization scope, file one document into multiple visible folders, share by user or role and return to earlier revisions when an authorized editor needs to restore content.

What is implemented
The documents routes and schema implement personal and organization ownership, folder collections, VIEW through MANAGE access levels, document versions, archive states, protected media and rotatable public document links.
Availability boundary
The documents feature and the person's personal or organization permissions determine creation and access. Folder membership, sharing, restore and public-link actions re-check the live access level.

Uploaded sources, collections and passage search

Upload the source files the organization relies on, extract and divide their text into traceable passages, group them into collections and search the indexed material before opening the source detail.

What is implemented
The knowledge workspace implements multi-file upload with indexing status, collection management, highlighted passage results, extracted-text inspection, reindexing and archive or delete actions. Postgres full-text search remains available even if best-effort embeddings fail.
Availability boundary
Knowledge access requires the knowledge feature and its view or mutation permission. Uploads obey the configured file-size rule and the plan's knowledge-source quota.

Knowledge answers constrained to indexed sources

Ask a question in the dedicated knowledge mode and inspect the source chips beneath the answer instead of receiving an unattributed response from general model knowledge.

What is implemented
The knowledge workspace displays cited source documents, and the API instructs knowledge mode to answer only from selected document excerpts, report when the material is insufficient and return the sources actually used.
Availability boundary
Knowledge chat requires knowledge:view_global, assistant:use_knowledge_chat, the knowledge feature, a usable configured AI provider and the applicable AI usage limits.

Role-defined authority and a permission-gated activity history

Configure role permissions for the work a person should perform, then let authorized administrators search the activity record by actor and action when they need to understand what changed.

What is implemented
The role editor exposes the permission catalog, while the activity page reads the agency audit feed with actor, action, time, detail and entity links behind audit:view_global.
Availability boundary
Custom roles, organization structure and audit visibility have separate features and permissions. Seeing the activity feed does not grant authority to repeat the action it records.

Previewed organization recovery with protected restore actions

Create an organization recovery point, inspect what it captured, preview the difference before a restore and require a stated reason plus typed organization confirmation for the consequential operation.

What is implemented
The recovery surface implements point creation, package inspection, restore preview, guarded confirmation, job history, rollback handling and optional scheduled or selective recovery declarations.
Availability boundary
Recovery points require the recovery_points feature and dedicated view, create or restore permissions. Selective scope and scheduling are optional plan capabilities with their own permission and retention rules.

Read before choosing

Knowledge becomes useful only when its source and authority remain clear

The product keeps authored documents, indexed sources, AI answers and recovery operations separate so one convenient surface cannot silently inherit another surface's authority.

  1. Authoring a document does not index it automatically

    The Documents workspace owns editable working material and its access history. The Knowledge Base owns uploaded, extracted and searchable sources. A document is not available to cited knowledge answers until it enters that separate indexing flow.

  2. Cited answers still require configured AI

    Full-text knowledge search works without a model provider. Conversational answers require a configured provider, specific permissions and usage capacity; knowledge mode must say when the indexed sources do not support an answer.

  3. Recovery is a governed product workflow, not unlimited backup authority

    Point count, retention, selective scope and scheduling depend on the live plan and configuration. Restore remains a separately permissioned, previewed operation and does not replace an organization's independent compliance or continuity requirements.

Next decision

Choose what you need to verify next

You do not need to read the whole site. Continue with the fit, adoption, price, or control question that matters to this decision.

  1. Run a structured CRM evaluation

    Name the broken handoff, map the complete workflow, and ask each provider for evidence before choosing.

    Use the evaluation guide
  2. Confirm plan fit and current price

    Review published plans, billing periods, included capabilities, limits, and configuration conditions.

    Review pricing
  3. Review control and risk questions

    Inspect isolation, permissions, approvals, audit evidence, continuity, and deletion boundaries.

    Open the trust center

Bring one decision that depends on a document only one person can find.

We will trace how the material is authored, shared, indexed, cited and protected through the real product boundaries.