Automate a weekly AWS cost report with anomaly flags

AWS shows spend in Cost Explorer, but the graph helps if someone looks at it. A forgotten instance, a runaway log group or a data transfer spike can run for weeks before the invoice reveals it. Looking weekly instead of monthly shortens that window, yet the weekly look is a chore that gets skipped. The job is simple enough to schedule: pull last week, compare it with the week before, and say what changed.

You add the job and grant view access to cost and usage data. Each Monday a scheduled agent requests the previous seven days grouped by service, using the Cost Explorer GetCostAndUsage operation, and compares each service with the prior week. It reports the total, the three biggest increases with the amount and percent change, and any service that appeared this week. If AWS Cost Anomaly Detection is enabled on your account, it also lists the anomalies that service has raised. It cannot change resources and it does not stop anything; it reports.

The CTO agent, as department head, reviews the draft before it reaches you: do the totals add up, are the percent changes computed against the right week, is any movement explained by something you already know, such as a release. Then you approve, edit or skip it, and nothing is sent until you do. Because the agent sees numbers and not intent, tell it about planned spend, like a load test, in the notes field so it stops flagging it.

Treat the top mover as a question, not a verdict. Open the service in Cost Explorer, filter by usage type, and find the resource. The report states how much spend changed, not whether that is acceptable, so judge it against your own budget. Over a month the same names will keep appearing; those are the places worth a cleanup.