Your CAB meets weekly and every production change needs approval. Lead time is 9 days, of which 6 are waiting. The board will not simply be removed. What is your approach?
Show the full answer Hide the answer
What the interviewer is testing
Whether you can change a governance process by working with its intent rather than by declaring it obsolete — which is the difference between an architect who influences and one who complains.
The approach
Do not attack the CAB. Its purpose is legitimate: someone is accountable for production stability and wants evidence that changes are controlled. The argument to win is not "we should skip approval" but "here is stronger evidence than a signature".
Use the standard change category. ITIL already contains it: a pre-authorised, low-risk, repeatable change type executed against a standing authorisation. This is not a workaround; it is the framework's own answer.
Build the evidence package, which is the actual work and is architecture work:
| CAB concern | Automated equivalent |
|---|---|
| Has this been tested? | Pipeline gates on test suite, coverage delta, contract verification |
| Is it secure? | SAST, dependency scan, secret scan, policy checks, all blocking |
| Is it authorised? | Peer-reviewed pull request, no standing production write access |
| Can we undo it? | Automated rollback, verified backward-compatible migrations |
| Will we know if it fails? | SLO monitoring, canary analysis with automatic abort |
| Is there an audit trail? | Immutable pipeline records linking commit, approver, artifact, deployment |
Pilot narrowly. Pick one low-risk service, agree with the change authority that deployments meeting the criteria proceed without review for three months, and report the outcome — deployment count, change failure rate, incidents. Evidence from your own organisation is far more persuasive than industry research.
Bring the DORA finding, carefully: teams deploying frequently in small batches have lower change failure rates, and approval by someone unfamiliar with the change correlates with worse outcomes. Present it as supporting evidence, not as an argument that the board has been wrong for a decade.
What a strong answer adds
Retaining a genuine human gate for the cases that warrant it — first-of-kind changes, irreversible migrations, changes to regulated flows — because proposing that everything becomes automatic invites rejection of the whole idea. Also: give the change authority better visibility than they have now, a dashboard of what deployed and what happened, which usually addresses the underlying anxiety.
Common weak answers
"CABs are proven ineffective, we should remove it." True in the research and a losing argument in the room. Automating the CAB's meeting rather than replacing its function.