Sample report — synthetic example, not your account
This is exactly what the agent hands you on day 15 — a ranked findings table on the left, a short voice memo on the right. Every number is fabricated for this page; your account would produce a different table with its own narrative.
| Resource | Issue | Monthly waste | Recommended action | Confidence |
|---|---|---|---|---|
| cloudwatch-logs-all-app | 1.8 TB in Standard, 38-day retention, mostly access logs | $2,140 | Move to Logs Infrequent Access, set 30-day retention | High |
| redshift-analytics-prod (dc2.large × 6) | CPU avg 9% across the cluster; queue depth near zero overnight | $1,946 | Consolidate to 3 dc2.large nodes, keep headroom for monthly report job | High |
| lambda-billing-processor | 14M invocations/mo at 1,536 MB memory — peak usage 380 MB | $1,224 | Right-size memory to 512 MB, batch invocations to reduce cold starts | Medium |
| cloudsentinel-prod-db (db.r6g.4xlarge) | CPU avg 6%, memory 22% over 30 days — heavily oversized | $1,180 | Downsize to db.r6g.xlarge, monitor for a week | High |
| elasticache-prod-cache (cache.r6g.4xlarge) | CPU avg 4%, zero evictions for 21 days | $1,180 | Downsize to cache.r6g.xlarge | High |
| i-0a2f8b1c9d3e4f5a6 (m5.2xlarge) | Idle 21 days — avg CPU 0.4%, no network connections | $412 | Snapshot, then stop the instance | High |
| NAT gateway (us-east-1c) | 1.1 TB/mo egress to S3 in the same region — cross-AZ, not cross-region | $148 | Add a VPC endpoint for S3 to keep traffic private | Medium |
| ebs-vol-03a4b5c6d7e8f9012 | 12,000 IOPS provisioned, peak usage 400 IOPS | $92 | Convert to gp3 and set IOPS to 1,000 | High |
| s3://logs-archive | 4.2 TB in Standard, 340-day average object age | $86 | Move to Glacier Instant Retrieval | High |
| eip-01a7b9c0d1e2f3a4b | Elastic IP unattached for 47 days | $4 | Release the address | High |
| Total estimated waste | $8,412/mo | 10 findings · reconciles to the headline | ||
From the agent
I sort rows by monthly waste first —$2,140 for CloudWatch logs is not the same kind of decision as $3.60 for an unattached Elastic IP, but the small ones are usually free wins so I keep them in the list rather than ignoring them. The two Medium-confidence rows are the ones I would want a human to confirm before I act; the rest is straightforward.
The us-east-1c NAT gateway is carrying 1.1 TB a month to S3, which is the same region. Every byte is paying the cross-AZ data-out rate, which is the single silliest line item on most AWS bills. A VPC endpoint for S3 would route this traffic inside the AWS network and drop the cost to roughly zero — I would do this one before trimming the cheaper rows below it.
The ebs-vol-03 row is the textbook win: provisioned IOPS you set during a launch three years ago and forgot about. gp3 lets you pay for exactly the IOPS you use, with 3,000 free per GB. The savings are small in absolute terms but the change is one API call and zero risk — the kind of finding I file automatically without asking.
If you give me write access to this account, I run the safe subset first: release the unattached Elastic IP, convert the EBS volume to gp3, and create the S3 VPC endpoint. I would batch the rest into a single change request so one human can review my plan before I touch the RDS instance or the ElastiCache node. The $8,412/mo figure is what I would expect if every row is acted on; the conservative subset clears roughly $2,400/mo on its own.
Your account
Connect in five minutes. The agent observes for two weeks, then delivers a report in this shape on your own infrastructure.
Want a real report on your account? →