Skip to main content
← Back to enterprise readiness

Technical review artifact

Technical review boundary mapping

A scope-capture worksheet for mapping which AI execution paths need governed authorization, which paths remain outside the guarantee, and what evidence a pilot must produce.

Download Markdown

Review context

Organization

Capture during technical review.

Review date

Capture during technical review.

Primary reviewer

Capture during technical review.

Security reviewer

Capture during technical review.

Platform owner

Capture during technical review.

Audit, risk, or compliance stakeholder

Capture during technical review.

Trigger for review

Capture during technical review.

Step 1

Highest-Scrutiny AI-Driven Action

Identify the action that would create the most scrutiny if it had to be explained three months from now.

Workflow name
AI-driven action
Business system affected
Customer, transaction, record, or operational impact
Why this action is high scrutiny
Existing control owner
Existing evidence source

Can the current system prove why this action was allowed before it ran?

Step 2

Execution Path Inventory

List every system that can initiate, transform, approve, or execute the action, then classify which paths are inside the governed boundary.

System or actor
Action
Execution-capable?
Current evidence
Inside governed boundary?
Required change

The enforcement guarantee applies only where execution-capable paths route through the governed boundary.

Step 3

Identity And Authorization

Capture the OIDC/JWT claims and role mappings required for a controlled pilot.

Identity provider
OIDC issuer
JWT audience
JWKS URL ownership
Required role claim
control_owner mapping
policy_author mapping
operator mapping

Execution control is not reviewable without a defined identity boundary.

Step 4

Policy Ownership

Document which policy authorizes the action, who owns it, and what evidence must prove evaluation before execution.

Policy or control
Accountable owner
Author
Approver
Review cadence
Evidence required

What happens if policy evaluation fails, and what denial record is preserved?

Step 5

Approval, Escalation, And Timeout Rules

Define which actions require approval, who can approve, how escalation works, and what happens on timeout.

Action class
Approval required?
Approver role
Escalation path
Timeout behavior
Self-approval blocked?

The approver cannot be the actor requesting the action. If approval, escalation, or evidence capture fails, execution must not proceed.

Step 6

Required Evidence Artifacts

Mark the artifacts required for audit, incident review, procurement review, or pilot acceptance.

Intent or request record
Policy decision record
Approval request record
Approval envelope or authorization record
Pre-execution evidence record
Execution record
Parameter comparison record
Denial record
Checkpoint or replay record
Hash or integrity evidence

Can a reviewer reconstruct request -> policy -> decision -> outcome without operator interpretation?

Step 7

Recovery, Retention, And Data Durability

Capture the durability standard for execution records and the restore validation expected by the buyer.

Backup interval
Restore validation cadence
Retention period for execution records
Retention period for denied actions
Evidence export format
Database ownership
Disaster recovery objective

Execution records only matter if they survive failure and can be restored without operator interpretation.

Step 8

CI, Security, And Procurement Evidence

List the evidence required by procurement or security before a pilot can be approved.

Enforcement test results
Auth test results
PostgreSQL integration test results
Backup and restore validation
SBOM
SAST output
Secret scanning output
Container scan output
Provenance or signing evidence
Threat model

Do not imply the current package proves more than the artifacts show.

Step 9

Pilot Scope

Convert the review into a deployment scope with explicit in-scope paths, out-of-boundary systems, and blockers.

In-scope workflow
Explicitly out-of-scope workflows
In-scope execution systems
Out-of-boundary systems
Required identity integration
Required policy definitions
Required evidence exports
Success criteria
Pilot blocker

The review output should state whether the selected path can prove control: yes, no, or partial.

Review output

The completed worksheet should state whether the selected execution path can prove control, what gaps remain in the current stack, and whether the pilot should proceed, revise scope, or wait.

Schedule technical review