A change advisory board meets weekly and approves about 60 changes in a 40-minute meeting. Attendance is good and the minutes are complete. What assurance does the meeting actually provide?

The arithmetic first

40 minutes across 60 changes is 40 seconds each, including the ones that generate discussion — which means most receive a few seconds. In that time a board can confirm that a change has a ticket, an owner and a stated window. It cannot assess whether the change is safe, because nobody in the room has read the diff.

That is not a criticism of the people. It is a capacity statement, and it is checkable from any organisation's own meeting minutes.

What the meeting does provide

Something real, and worth naming so the conversation is not a strawman: scheduling coordination and a record. The board catches two teams deploying into the same window, a change landing during a trading freeze, and an owner who does not know a dependent system exists. Those are genuine, and they are coordination functions rather than risk assessment.

Why the other options fail

• "Strong assurance because every change was reviewed." This conflates a formal step with an effective one, which is exactly the substitution the arithmetic exposes. The record is complete and the review is nominal.
• "None at all — advisory boards never help." Overcorrection. Published research on delivery performance consistently finds that heavyweight external approval does not improve change failure rates and does slow delivery, but coordination and record-keeping still have value; the honest claim is about risk assessment specifically.
• "Proportional to seniority." Seniority does not create reading time. A more senior board approves 60 unread changes with more authority.

What to do instead

Move the technical assessment to where the information is — peer review on the change itself, automated policy checks, and a deployment pipeline that enforces staged rollout and automatic rollback. Then reserve the board for the small number of changes that genuinely need cross-team coordination: those touching shared infrastructure, those inside a freeze, and those whose rollback is not automatic.

The measurable claim to make when proposing this: change failure rate and time to restore should not worsen, and both should be instrumented before the change so the argument is settled by data rather than by seniority.

When this is the wrong answer

Where a regulator or a contract specifies a named approver for a class of change, the board is an obligation rather than a design choice, and the work is to narrow its scope rather than remove it. Keep it for the changes the obligation names, automate the rest, and be able to show the split.

It also stays right for genuinely irreversible changes — a migration that destroys the old store, a regulatory filing, a change to money movement — where slow deliberate approval is proportionate precisely because rollback is not available. The rule is reversibility, not risk: choose deliberate approval when rollback is unavailable, and automated gates plus staged rollout when it is.

What the automated path costs, and should be stated when proposing it: policy rules to maintain, a staged-rollout capability that has to be built, and the loss of the scheduling conversation, which must be replaced by a calendar the board used to hold in its head. Research programmes publishing on delivery performance since 2014 have consistently found the throughput benefit real and the coordination loss underestimated.