Unclassified

Governance & Approval

Governance is the hard boundary between proposal and execution. IntegrateWise does not allow consequential actions to bypass governance.

For the canonical governance model, see docs/spec/INTEGRATEWISE_CANONICAL_SPEC_v1.0.md, docs/architecture/UNIFIED_GOVERNANCE_SPEC.md, and docs/platform-specs/09-governance-engine.md.

Two gates

Governance uses two phases:

Pre-Proposal Gate

Evaluates:

  • authority
  • scope
  • policy

Possible results:

  • ALLOW_PROCEED
  • REQUIRE_PRE_APPROVAL
  • DENY

A 0.99 confidence proposal with missing authority is still DENY.

Post-Proposal Gate

Evaluates:

  • risk
  • evidence
  • confidence
  • approval requirement

Possible results:

  • AWAITING_HUMAN_APPROVAL
  • HUMAN_APPROVED
  • DENY

Confidence never grants authority. High confidence does not remove approval requirements for consequential actions automatically.

Approval Center

The Approval Center is the user-facing governance surface. Every proposed consequential action must flow through it.

Approval records include:

  • proposed action
  • business purpose
  • initiating actor
  • affected contexts and systems
  • exact fields or operations
  • evidence
  • execution-plan version
  • risk classification
  • approval policy
  • expected outcome
  • rollback or recovery availability

Available decisions:

  • Approve
  • Reject
  • Request Changes
  • Edit and Approve when policy permits it
  • Delegate Approval
  • Cancel

ExecutionPlan contract

The user never approves an AI message. The user approves an ExecutionPlan.

ExecutionPlan contains:

  • plan identity and tenant scope
  • proposal lineage
  • immutable plan version
  • summary and impact summary
  • step dependencies and compensation steps
  • affected entities and systems
  • authority and evidence references
  • execution mode and sync mode
  • rollback strategy

After approval, the Workbench shows execution progress, provider responses, partial completion states, failure and retry states, sync results, and final outcome.

Governing rules

  • Approval Center is the only surface that can mint an approval token.
  • No handoff package can be generated without approval lineage.
  • Twin does not auto-execute.
  • Governance rules are compiled configurations. They cannot be modified by model outputs or agent workflows.

Audit expectations

Every consequential action must log:

  • actor
  • authority scope
  • policy validation result
  • affected context
  • exact provider operations
  • execution-plan version
  • approval record
  • provider outcome
  • sync result
  • tamper-proof audit event

This is not optional in production integrations.