Skip to main content
Safe autonomy requires separation of concerns. IntegrateWise divides the world into two layers: canonical and operational. The canonical layer holds the authoritative business state. The operational layer is where AI proposes, experiments, and acts. Governance sits between them, ensuring nothing reaches the canonical record without authorization, scope, and a receipt. This separation is what makes it possible to give AI broad freedom to work without risking the ground truth the business depends on.

​
Two Layers, One Purpose

Canonical Layer

The authoritative business state. Reflects source systems and the unified context Spine Fabric assembles from them. This is the durable memory of the business: entities, events, evidence, decisions, and receipts. AI does not write here directly.

Operational Layer

The agentic workspace where AI drafts, proposes, simulates, and executes. Mistakes are recoverable. Proposals can be rejected. Nothing here is authoritative until it passes through governance into the canonical layer.

​
What Separates Them

Governance is the bridge and the gate. It enforces authorization, scope, approvals, and receipts for every operation that moves from operational to canonical. This is not a configuration setting. It is a structural property of how Spine Fabric and SpineWorkspace are built.

​
The Governed Path From Operational to Canonical

Every meaningful action follows the same path. This path keeps the canonical layer clean while giving AI and humans maximum flexibility in the operational layer.
1

Propose

The agent or human drafts an action in the operational layer: a message, a record update, a workflow. It is assembled from shared context but has not yet touched the canonical state.
2

Simulate

Where possible, the effect is previewed. The agent shows what would change, which systems would be touched, and what the outcome would look like. The human reviews with full transparency.
3

Approve

Governance applies authorization checks, scope validation, and approval rules. The human with authority confirms. The capability is unlocked.
4

Write-back

The governed capability executes against the relevant source systems. The canonical layer updates to reflect the new truth. The change is permanent and auditable.
5

Receipt

A receipt is generated and attached to the shared thread. The business memory now includes what happened, who approved it, and why. The next agent or human starts from this updated state.

​
Why This Separation Matters

Without it, AI autonomy becomes a liability. Agents that write directly to systems of record can corrupt data, lose audit trails, and create competing versions of truth. With it, agents have broad latitude to work inside the operational layer while the canonical layer remains protected. This design also makes trust practical. Humans can see what the agent proposed before any system changes. They can reject without damage. They can approve with confidence because the simulation showed the effect. Over time, pre-approved scopes can expand for reliable agents, but the governance boundary never disappears.
The operational layer is not a sandbox. It is a real workspace where real work happens. Proposals, drafts, and simulated outcomes are visible, shareable, and part of the continuity loop. Only the write to canonical state is gated.

​
How Spine Fabric Implements This

Spine Fabric maintains the canonical layer as a unified, governed continuity record. The Spine Agent Network operates in the operational layer, proposing and preparing actions. SpineWorkspace displays both layers to the human user, with clear visual distinction between what is proposed and what is authoritative. Governance capabilities in the fabric enforce the boundary, so no agent, no matter how capable, can bypass the approval path to canonical state. This separation is not overhead. It is the architecture that makes safe autonomy possible at scale.