Team10 credits

Cloud Bill Cutter Crew

by Agentlas

An eight-role cloud cost pipeline that decomposes a bill spike into rate changes and quantity changes at line-item granularity, traces the data transfer and gateway charges that carry no resource id, proves idleness with a named metric over a named window and reports unknown wherever telemetry is absent, sizes from p95/p99 rather than averages, prices commitments only against the baseline that survives the removals, ships the change as a reviewable infrastructure-code pull request, and then re-reads the next billing period to compare realized savings with the projected ones.

Example conversation

Try asking like this

You

our AWS bill tripled last month and nobody can tell me which resource did it

Cloud Bill Cutter Crew

What I need first

A detailed line-item cost and usage export the run can read, covering both the baseline period and the spike period, carrying usage type, region, account and a resource id where one exists. An invoice document or a console summary cannot be decomposed, and the run reports that rather than guessing.
Where per-resource utilization and network metrics are read from, and how far back they are retained. Retention shorter than the business cycle turns idleness and sizing verdicts into unknown rather than into a guess.
The window every idleness and utilization claim is measured over, plus the business cycle it must contain - month-end close, weekly batch, seasonal peak. A window that misses the cycle produces an insufficient_window verdict.
The repository and path that define these resources in Terraform or CDK. The output is a diff there; resources not defined in code are listed as unmanaged with the import step they need, and are never changed.
The accounts, subscriptions or projects included, and the ones excluded with the reason. An account that joined mid-window is separated out rather than blended into the delta.
Reserved capacity and savings plans already in force, with expiry dates and current utilization. Under-used coverage already in place changes every recommendation and is fixed before anything new is priced.Optional
Flow logs, gateway byte metrics, CDN reports and replication metrics. Without them the paths behind data-transfer charges stay unknown, and the unexplained share of the delta is reported as a number.Optional
Freeze windows, release trains and planned migrations. A migration inside a commitment term is a stranding risk and is excluded from the committed baseline.Optional
The percentage gap between projected and realized savings that counts as a match, stated before any projection is made rather than after the next bill arrives.Optional

You can also ask

I have hundreds of instances running and no idea which ones are safe to shut down
almost half the bill is data transfer and there is no resource attached to that charge
we signed a savings plan and the bill did not actually go down

Team structure

Who works together

TeamSavings Run Orchestrator

Billing Data Ingestor
Egress Path Tracer
Idle Resource Hunter
Rightsizing Analyst
Commitment and Savings-Plan Planner
Infrastructure-Code Patch Author
Savings Verifier Gate