Can save
How Can save /mo works
Can save is the estimated monthly amount you could stop paying if you fix open waste — not a refund of spend already billed this month. The total rolls up Cloud Cost Management findings (AWS, Azure, GCP) and AI Cost Management findings (OpenAI / Anthropic / Gemini keys, plus cloud-native AI lines when those accounts are connected).
The formula
Each waste finding stores an estimated monthlySavings from the scanner (retail / list-price heuristics — e.g. unattached disk $/GiB-mo, empty APIM SKU, idle static IP, idle API keys, AI spend anomalies).
where status ≠ rejected (cloud + AI findings, same pool)
Amounts are converted to your org display currency. Rejected findings are excluded so dismissed noise does not inflate the total. You do not need both products subscribed for the math to work — connected sources that produced findings simply contribute.
Can save vs Cost vs Verified
| Column | Means |
|---|---|
| Cost (MTD) | What the service / org has already been billed this month (synced from Cost Explorer / Cost Management / billing export). |
| Can save /mo | Estimated ongoing monthly spend you avoid from now on if you fix open cloud and/or AI waste. Not a credit for past Cost. |
| Verified | Confirmed after you fixed something and spend drop was observed (observedSavings when available). |
On Savings → Cost by month
- Current month (live) — sum of monthlySavings for all non-rejected actions right now (open, approved, in progress, fixed, and verified).
- Past months — same sum, but only actions whose createdAt falls in that calendar month (findings opened then — not a frozen end-of-month snapshot).
- Verified on that table is attributed by verifiedAt month.
On Services
Per service row, Can save /mo is the sum of monthlySavings for actions mapped to that service (AWS CE / Azure CM / GCP billing bucket, OpenAI / Anthropic / Gemini provider rows, or waste-type → service mapping). Same exclude-rejected rule.
Where estimates come from
Scanners use public list / retail heuristics and inventory or usage signals (Resource Graph, Compute APIs, CloudWatch, provider usage APIs, etc.). They prioritize action — they are not your exact invoice line. Cloud-native AI meters (Bedrock, Vertex, Azure OpenAI) stay under Cloud waste scans; OpenAI / Anthropic / Gemini API-key waste under AI waste scans. Both feed Can save.