The front door
Cloud + AI Cost and Reliability Assessment
A ranked backlog of cost reductions priced from your own billing data, with effort, risk, and a verification method on every line, plus a map of where your infrastructure fails under growth. If nothing in the backlog is worth at least the fee, that will be visible in your own numbers, not my claims.
Six deliverables, one written report.
Spend baseline
13-month trend and current-quarter decomposition by service, account, region, tag/team, and usage type, computed from your Cost Explorer and CUR data. Any unexplained bill jump gets decomposed to its cause.
Ranked savings backlog
Target 15 to 25 items. Each carries: the finding, its evidence (billing lines and resource IDs), first-year dollars, effort in engineer-days, production-risk score, confidence, suggested owner, rollback note, and the named verification method: which billing line moves, after what lag.
Commitment posture review
Savings Plans and Reserved Instance coverage and utilization, expiring commitments, and renewal risk. Analysis for your finance and engineering decision; I never purchase commitments on your behalf.
AI workload cost attribution
Where present: Bedrock, SageMaker, GPU instances, inference infrastructure. Spend by workload, team, and feature, unit-cost trend, and the industry overspend patterns checked against your actuals.
Failure-surface map
Where the infrastructure breaks under growth or partial failure: single-AZ exposure, missing alarms on revenue paths, backup and DR gaps, quota ceilings, unowned critical resources. Each tied to a business consequence.
Implementation menu and walkthrough
Fixed-price sprint and retainer options mapped to your backlog clusters and their identified dollars, plus a 45-minute walkthrough call. The report is deliberately complete enough that your team can ship it without me.
Fixed. Visible. Never hourly.
-
$5,000 to $7,500
Standard engagement
Single account or organization under ~$75k/month quotes toward the floor; multi-account orgs, $75k to $200k/month, or heavy AI spend quote toward the top of the band.
-
$10,000 to $15,000
$200k+/month AWS spend
Larger orgs are more work and carry more identified dollars; the band scales accordingly.
The credit rule
The assessment fee is credited in full against any implementation engagement over $5k, valid 90 days from report delivery. Payment is 100% on signature; there is no milestone billing on a fixed ten-day deliverable.
Why the price is what it is: the fee works out to roughly 3 to 5 percent of identified first-year savings at typical waste rates. Industry surveys put cloud waste at 27 to 29 percent of spend; at $50k/month, run that arithmetic on your own number and the fee justifies itself before the credit rule even applies.
What this is not.
Scope boundaries stated up front, because they're also the risk controls.
- xNo production changes of any kind. The assessment is read-only, full stop. Implementation happens in a sprint, with written approval and a rollback plan on every change.
- xNo guaranteed savings. Identified savings are priced from billing data; realized savings require implementation and are verified from CUR after the billing lag.
- xNo AWS resale, credits sourcing, or managed billing. I am not a reseller and have no incentive tied to your spend going anywhere but down.
- xNo EDP or private-pricing negotiation.
- xNo security audit beyond cost-related and reliability-related risk notes.
- xNo access to customer data, application secrets, or source code. IaC repo read access is optional and separately consented. A data-handling checklist is countersigned before any access is granted, and access revocation is confirmed in writing at the end.
The ones I actually get.
AWS already shows us recommendations. Why pay for this?
Correct, and that is exactly the gap. Recommendations without ownership, risk scoring, rollback plans, and verification do not get merged. The backlog you get is the layer between the dashboard and an approved PR.
Honest test: when did your team last ship one of those recommendations?
We already have a FinOps tool.
Is it getting recommendations merged and verified weekly? If yes, you genuinely don't need me, and I'll say so. Usually the answer is no, because tools produce findings and findings aren't changes. The assessment prices, ranks, and risk-scores the work so it actually clears your engineering bar.
Can you guarantee savings?
No, and be suspicious of anyone who will. Savings are identified from your billing data and verified from CUR after implementation. What I can state plainly: if nothing in the backlog is worth at least the assessment fee, that will be visible in your own numbers. The credit rule means the assessment costs nothing if you proceed to implementation.
Why not just hire a FinOps person?
A FinOps hire is $150k+ a year and three to six months to productivity. The assessment is a fixed fee and ten days, and if you hire anyway, the backlog is the new hire's onboarding doc.
How do I know the numbers are real?
Every dollar figure traces to deterministic computation over your exported billing data: scripts and saved outputs, not model estimation, and rounded down. Each backlog item names its evidence (billing lines, resource IDs) and its verification method (which CUR line moves, after what lag). The report is built to survive being forwarded to your sharpest engineer.
What do you need from us?
A signed order form, a countersigned data-handling checklist, read-only access via a provided IAM role template, and a written intake questionnaire. Total effort on your side is an hour or two, most of it for whoever grants IAM access. The ten-day clock starts at access grant, not signature.
