Before a cloud workload goes live, name the people who will own access, monitoring, and recovery. A short review with those owners can reveal gaps while the team still has time to fix them.
Who owns each control?
Start with the provider’s responsibility model for the services you use. On AWS, for example, running software on a virtual machine leaves the customer responsible for patching its guest operating system. A managed service can shift some of that work to the provider. Record who owns each remaining control and who can verify it. AWS shared responsibility model
Who can access production?
Review human and application access separately. Give each application the permissions it needs for its job, and review how administrators get elevated access. Include a way to revoke access quickly if a credential is exposed. AWS identity and access security guidance
Will someone notice a risky change?
Identify the events that deserve attention, such as a change to a privileged role or a new public endpoint. Send alerts to an owner who can investigate. During a review, make a safe test change and check that the alert reaches the right person. AWS Well-Architected Security Pillar
Can the team recover?
Pick a workload and run a restore in a separate environment. Measure how long recovery takes, check whether the restored data is usable, and record which steps depend on a particular person or credential. Use what you learn to update the recovery plan before launch.
Worked example — A customer portal moves to the cloud
Imagine a team moving its customer portal from an office server to a cloud environment. The platform team owns configuration and backups, the application team owns its code and dependencies, and the security team monitors changes to privileged access. Before launch, the teams test a restore in a separate environment and confirm who responds if an alert fires outside business hours.
Before launch
- Name an owner for each control and verify which responsibilities sit with the provider.
- Review production access for people and applications, including how to revoke it.
- Trigger a safe test alert and confirm it reaches someone who can respond.
- Restore a backup and record how long recovery took and whether the data was usable.