Review this design authority. Every change above £50,000 or touching customer data goes to a board of seven that meets fortnightly. Lead time to a decision averages three weeks. Last year it reviewed 94 submissions and rejected two. Teams have started splitting work to stay under the threshold. What would you change, and what would you keep?

What the numbers already tell you

A 2% rejection rate over 94 reviews is not evidence that the estate is healthy. It means one of two things: the board is approving things it should question, or it is reviewing things that never needed a board. The splitting behaviour settles it — teams are routing around the control, which means the threshold is driving behaviour and the behaviour is not the one intended.

Three weeks of lead time on a fortnightly meeting is mostly waiting, not reviewing.

What I would remove

• The fixed monetary threshold. It is gameable by construction and correlates weakly with architectural risk. A £30,000 change introducing a second identity provider matters more than a £200,000 hardware refresh.
• The fortnightly meeting as the default path. Most submissions need one competent reviewer and an hour, not seven people and two weeks. Asynchronous review with a named reviewer clears the majority.
• Review of anything already covered by an automated check. If a policy engine can verify encryption, region and tagging, the board reading a diagram to confirm them is theatre.

What I would replace it with

Trigger on properties, not price. A short list: a new data classification crossing a boundary, a new third party processing customer data, a change to an authentication path, a new persistence technology, anything that creates a single point of failure for more than one product. Those genuinely need the room.

Then split the function in two: a fast advisory path, available on demand, that most teams use voluntarily because it is useful; and a small blocking path for the trigger list. Advice that is fast gets asked for; advice that costs three weeks gets avoided.

What I would keep, even though it looks like part of the problem

• The board itself, for the trigger list. Some decisions genuinely benefit from seven perspectives, and removing the forum entirely is the overcorrection that follows this critique about a year later.
• The written record. Its value is not the approval, it is that a decision and its reasoning exist in 2029 when nobody involved is still there.

When not to change any of this

If the board exists because a regulator requires a named approval body for material change, then the meeting is an obligation and the right move is narrower: keep the board, cut its scope to the regulated triggers, and put everything else on the advisory path. Check which of the two you have before proposing a redesign, because arguing the wrong one loses the room and the change.