Set up, understand, and operate Venturi

Venturi is the attribution layer for AI. It links every AI inference across providers, services, code owners, identities, and budgets into one attribution graph from evidence the enterprise already produces. These docs cover connecting evidence sources, validating attribution coverage, and operating Venturi across engineering, finance, security, and product workflows.

Start with a task

Understand the system

1. Evidence sources Cloud, identity, code, billing, telemetry, and API inputs.
2. IDD classification AI Inference Detection and Invocation Discovery (IDD) classifies a normalized observation before it can become an invocation. Discovery-only evidence remains an observation; it is not an invocation.
3. RAIL attribution The Reconciliation Attribution Intelligence Layer resolves ownership, confidence, output state, and evidence.
4. Allocation and customer decisions Teams inspect, reconcile, allocate, export, dispute, or accept the result.

IDD decides what a source record establishes before RAIL attributes a canonical invocation. This boundary prevents discovery signals, aggregate billing data, or network observations from being presented as request-level invocation truth.

Follow your role

Read product state consistently

Use Docs authority and product state to verify a capability against the customer agreement, deployed configuration, and tenant-specific evidence. These labels keep intended behavior separate from verified availability:

Current

Verified for the tenant's deployed configuration.

Preview

Available for evaluation, but not part of the onboarding baseline.

Target state

The intended customer contract; verify prerequisites and evidence before treating it as deployed.

Milestone

A named delivery checkpoint, not evidence of tenant availability.

Need help with an integration or operating question? Review Support.