Evidence Based Approval
Replacing a human judgement about whether a change is safe with a machine-produced record of the checks it passed, assessed once for the class rather than per instance.
A change advisory board reviews individual changes. At any real delivery rate it cannot review them meaningfully, so it approves on the strength of a description — which is a judgement about the quality of the description rather than of the change.
Evidence-based approval moves the assessment. The class of change is assessed once: what does the pipeline verify, what evidence does each run produce, how is rollout controlled, what triggers automatic rollback. Once agreed, an instance that produces conforming evidence is approved by that fact.
The evidence a run should carry: which commit, which tests passed, which scans and their findings, which approvals occurred, which environments it traversed, how it was rolled out and what the automated verification observed.
This is usually a stronger control than the board it replaces, and saying so is the route to getting it accepted. A board approving forty changes in an hour is providing a record of consideration; a gate that blocks on an unsigned artifact, a failed contract test or a canary breach is preventing specific failures.
What the board retains is the genuinely exceptional: changes outside the assessed class, changes during a freeze, and anything with a business impact that is not a technical question.