Waste Findings Categories and Priorities
Who this is for
Anyone who wants to reduce cloud spend by identifying and eliminating wasted resources.
What you will complete
Understand each category of waste CloudAIPilot detects, how findings are rated by confidence, and how to act on them.
Before you begin
- FinOps billing data must be connected.
- Navigate to FinOps in the left sidebar and scroll to the Potential Savings card.
Waste categories
CloudAIPilot detects twelve categories of cloud waste. Use the category filter dropdown in the Potential Savings card to focus on one type at a time.
1. Idle compute
What it is: A server that has been running but shows near-zero CPU, network, and request activity for an extended period. Likely a forgotten development or testing server.
How to act: Confirm no active workloads depend on it. Stop or terminate if confirmed idle.
2. Over-provisioned
What it is: A server that is consistently using only a small fraction of its allocated CPU or memory — paying for capacity that goes unused.
How to act: Resize to a smaller instance type (see KB-08-09 for the approval flow).
3. Orphaned resource
What it is: A cloud resource (volume, disk, snapshot) that is no longer attached to or associated with any active server but is still incurring storage charges.
How to act: Verify no server needs the resource. Take a snapshot for safety if needed, then delete.
4. Region drift
What it is: A server or resource running in a cloud region that is significantly more expensive than equivalent regions. Some regions cost 20–40% more for the same instance type.
How to act: Verify latency and compliance requirements. If a cheaper region is acceptable, plan a migration.
5. Dev/staging weekend
What it is: Development or staging servers that run 24/7 but are only used during business hours. Weekend and overnight compute is billed but idle.
How to act: Schedule automatic shutdown outside business hours using the AI Pilot or your cloud provider's scheduler.
6. Stale snapshot
What it is: Old backup snapshots that exceed your retention policy or belong to servers that have been deleted.
How to act: Review snapshot age and source. Delete snapshots outside your retention window.
7. Old-gen instance
What it is: Servers running on older-generation instance types where newer generations offer better performance at the same or lower price.
How to act: Migrate to the equivalent current-generation instance type during a low-traffic window.
8. Unused commitment
What it is: Reserved Instances (AWS), Committed Use Discounts (GCP), or Reserved VM Instances (Azure) that were purchased but are not being utilized, meaning you are paying for reserved capacity without using it.
How to act: Reallocate workloads to use the committed capacity, or allow the commitment to expire.
9. Cross-AZ transfer
What it is: High data transfer costs from traffic crossing availability zone boundaries within the same region. This often indicates a architecture issue where services are placed in different AZs unnecessarily.
How to act: Consolidate services into the same AZ where latency requirements allow, or use AZ-local endpoints.
10. Idle managed DB
What it is: A managed database instance (RDS, Cloud SQL, Azure Database) with near-zero query activity — running but not serving any application traffic.
How to act: Stop or delete the database if unused. Consider a serverless database tier for infrequent workloads.
11. Unused public IP
What it is: A reserved static/elastic IP address that is not attached to any running server. Most cloud providers charge for unattached reserved IPs.
How to act: Release the IP in your cloud provider console, or attach it to a server that needs a static address.
12. Stale bucket
What it is: An object storage bucket with little or no access over an extended period, still incurring storage charges.
How to act: Review bucket contents. Archive to cheaper storage tiers (e.g., Glacier, Coldline) or delete if no longer needed.
Confidence levels
Each waste finding is rated by confidence — how certain the system is that the resource is genuinely wasteful:
| Confidence | Meaning |
|---|---|
| High | Strong evidence of waste — resource is clearly idle or mismatched |
| Medium | Likely waste — usage pattern suggests underuse but has some variance |
| Low | Possible waste — ambiguous pattern; review before acting |
Start with high-confidence findings — they carry the most certainty and deliver savings with the least risk.
Acting on a finding
Each finding in the Potential Savings card has three actions:
- Acknowledge: Marks the finding as reviewed. It stays visible but is noted as seen.
- Snooze: Hides the finding for 30 days. Use when the resource is temporarily needed but you plan to act later.
- Dismiss: Permanently removes the finding. Use when you have reviewed it and decided no action is needed.
- Ask AI Pilot: Opens the AI Pilot with context pre-loaded so you can ask the agent to help execute the remediation.
You can also click Run scan now in the card header to trigger a fresh waste detection pass rather than waiting for the nightly automated scan.
To enable or disable specific waste categories, click Configure in the card header.