AI Computer Control
AI actions with operational guardrails.
AI Computer Control extends the assistant beyond analysis into controlled endpoint operations. It can execute defined tool actions for commands, scripts, services, file operations, disk workflows, and discovery tasks.
Execution Model
Every action is evaluated by the same AI risk system used across Breeze:
- lower-risk read operations can run automatically
- higher-risk mutating actions are gated by approval
- blocked operations never execute
This keeps AI useful for real operations while preserving change control.
Guardrails in Practice
AI tool calls are constrained by:
- role and permission scope (RBAC)
- tenant boundaries
- per-tool risk tiering
- approval workflow for sensitive actions
- audit logging for execution and denials
The model is action-level, so different operations within the same tool can have different enforcement behavior.
Operational Use Cases
Teams commonly use AI Computer Control to:
- run guided remediations across selected devices
- execute approved scripts faster during incidents
- perform safe file and service workflows
- trigger discovery or scan tasks with explicit oversight
Why It Matters
This feature turns AI from a chat-only surface into an operational copilot while ensuring technicians keep final authority over high-impact changes.
Capabilities
Device Tool Actions
Execute commands, scripts, services, file operations, and discovery tasks through the AI execution layer.
Risk-Tiered Enforcement
Every action is evaluated by the AI Risk Engine before execution; higher-risk actions require explicit approval.
RBAC Scope Control
Tool call scope is bounded by role, permission, and tenant context regardless of AI-initiated origin.
Execution and Denial Logging
All AI tool invocations and denials are recorded for post-incident review and operational audit.