Rules & Governance

Rules & Governance

• Context rules → Clarify whose context an AI Employee runs under.

• System rulebook → Permissions, scope, and operating modes.

• Limitations → Boundaries and unsupported scenarios.

---

Context Rules

AI Employees act within a specific context when performing actions. Context determines whose permissions, data access, and integrations are applied.

Rule	Details
Team-level execution	AI Employees run as team-level users, not as individuals. Their actions follow workspace-wide RBAC (role-based access control).
User-triggered context	When a user prompts an AI Employee in chat or assigns it a task, the action runs under the workspace’s permissions — not the individual user’s personal context.
Integration context	For third-party apps (e.g., Google Calendar, Slack), AI Employees use the configured connection associated with the workspace or skill. They do not access personal user accounts unless explicitly linked.
Approval context	Write operations (e.g., editing, creating, deleting) respect workspace-level permissions. If an action exceeds allowed scope, it is blocked or requires admin approval.
Scope inheritance	When running inside a project, task, or doc, the AI Employee inherits that object’s visibility and permissions. It cannot act outside the scope of the container unless separately authorized.

System Rulebook

The system rulebook defines the operating boundaries of AI Employees — including their permissions, scope of action, and supported modes of operation.

Rule	Details
Permissions	AI Employees follow workspace RBAC (role-based access control) just like human users. Their access level (Viewer, Editor, Admin) determines what actions they can perform.
Scope	AI Employees can only act within the workspace and projects they are added to. They cannot cross into other workspaces unless explicitly shared.
Modes of operation	Manual mode: Triggered directly by user input (chat or task assignment). Autonomous mode: Runs skills automatically on conditions or schedules.
Approvals	Sensitive or write operations always require explicit approval, either in chat or during skill execution.
Integration behavior	When connected to third-party services, AI Employees act only through the configured workspace connection.
Auditability	All actions are logged in workspace activity or chat history, including approvals, declines, and errors.

Limitations

AI Employees have defined boundaries that limit what they can do inside and outside of Motion.

Category	Limitation
Permissions	AI Employees cannot bypass workspace RBAC. If a human user role cannot perform an action, the AI Employee cannot either.
Connections	They cannot access external services without an authenticated workspace-level integration. Missing or revoked connections cause dependent skills to fail.
Context	AI Employees act under workspace context only — they cannot act under an individual user’s private data or personal accounts.
Data formats	Images, audio, video, and binary file contents are not processed. Only metadata is referenced.
Block syncing	No support for syncing blocks across multiple docs or projects. Copies are independent.
Autonomous scope	Autonomous skills are limited to the triggers and permissions configured. They do not run outside the defined scope.
Publishing	AI Employees cannot publish Docs or share external links without explicit approval.