Your architecture review board is a three-week queue. Half the submissions are checking things like region, encryption and tagging. Redesign the process.
Show the full answer Hide the answer
What the interviewer is testing
Whether you can separate what a machine should verify from what genuinely needs human judgement.
The diagnosis
The board is spending its scarcest resource — senior attention — on rule checking, which machines do better: consistently, instantly, and at 3 AM on a Friday.
That leaves no capacity for the work only the board can do, which is judgement about boundaries, trade-offs, sequencing and whether a design fits the business constraint.
The redesign
Move every rule-shaped check into automation:
| Currently reviewed | Becomes |
|---|---|
| Approved region | Organisation-level policy — non-compliant deployment is impossible |
| Encryption at rest | Policy as code at admission, plus continuous compliance check |
| Tagging and cost centre | Enforced at resource creation |
| Approved technology | Policy check against the radar's Adopt/Hold list |
| Network exposure | Default-deny with explicit, reviewed exceptions |
Note these become guardrails, not automated gates where possible — the unsafe action should be impossible or auto-corrected rather than caught after the fact.
Tier what remains by consequence and reversibility:
- Reversible and local: no review, team decides, record an ADR
- Hard to reverse or crosses boundaries: lightweight async review — an ADR circulated with a comment period, no meeting
- Commits the organisation for years or carries regulatory weight: the board, which now has time
The expected outcome
Most submissions disappear. The board meets less often, with fewer items, and the items are ones where senior judgement changes the answer.
What a strong answer adds
Running the new policies in warn mode first and fixing the existing backlog before enforcement, or the first day of enforcement blocks every team and the policies get exempted.
And giving teams a self-service pre-check so they can validate their design against the automated rules before submitting anything — which removes the review round trip entirely for compliant designs.
Common weak answers
Adding reviewers, which scales linearly against a growing queue. Removing the board, which loses the judgement function.