Cloud Cost Reviewer

• Executive summary — 5 lines, finance-readable, includes month-over-month and year-over-year deltas.
• KPI dashboard — total spend, vs budget, vs forecast, top 3 movers, cost-per-unit metric.
• Per-service breakdown — top 10 services by spend, each with %change vs prior period.
• Anomaly findings — sudden cost jumps, unused resources, pricing changes, untagged spend.
• Optimisation backlog — prioritised actions with estimated savings, owner suggestion, effort estimate.
• Reservation / Savings Plan utilisation — current commitment coverage, expiring soon.
• Cost-per-team / cost-per-application — if tags are clean, this is the most useful section for chargeback.
• Forward-looking note — flagged risks for next month, planned changes that affect cost.

Clarifying questions I will ask you

1. Which cloud(s) and SaaS providers? (AWS / Azure / GCP / M365 / Google Workspace / specific SaaS)
2. Reporting period? (Calendar month, fiscal month, custom)
3. Total spend this period? (Single figure, all sources combined)
4. Last period total + 12 months ago total? (For deltas)
5. Annual budget for this scope? (Drives "% of budget consumed" framing)
6. Top movers — which services / accounts / subscriptions had biggest swings?
7. Tag hygiene — are workloads tagged with Environment / Owner / CostCentre / Application? (Drives chargeback fidelity)
8. Reservations / Savings Plans / Committed Use Discounts — what's currently committed and when does it expire?
9. Anomalies you've already spotted? (Cross-check)
10. Audience for the report? (Engineering only / engineering + finance / exec sponsor / customer in MSP context)
11. Any planned changes next period? (Migration, new product launch, decommissioning)
12. Currency? (For consistent reporting)

Output template

# Cloud Cost Report — <period> — YYYY-MM

**Report ID:** CCR-<YYYY-MM>
**Period:** YYYY-MM-01 to YYYY-MM-DD
**Prepared by:** <name>
**Audience:** <internal eng / finance / exec / customer>
**Currency:** GBP / USD / EUR

## 1. Executive Summary
> <Plain-English, 5 lines. State total, vs budget, biggest mover, top 1 action, forward risk.>

## 2. Headline KPIs
| KPI | This period | Last period | Change | YoY |
|---|---|---|---|---|
| Total spend | <amount> | <amount> | +/- N% | +/- N% |
| Budget consumed | XX% | XX% | — | — |
| Forecast for full year | <amount> | — | vs <annual budget> | — |
| Cost per <unit> (e.g. monthly active user) | <amount> | <amount> | +/- N% | — |
| Reservation / Savings coverage | XX% | XX% | +/- pp | — |

## 3. Spend by Provider
| Provider | This period | % of total | vs last period |
|---|---|---|---|
| AWS | <amount> | XX% | +/- N% |
| Azure | <amount> | XX% | +/- N% |
| Microsoft 365 | <amount> | XX% | +/- N% |
| <SaaS> | <amount> | XX% | +/- N% |
| **Total** | **<amount>** | 100% | +/- N% |

## 4. Top 10 Services (by spend)
| Rank | Service | This period | % of total | Change | Notes |
|---|---|---|---|---|---|
| 1 | <service> | <amount> | XX% | +/- N% | <one-liner> |
| 2 | ... |  |  |  |  |
| ... |  |  |  |  |  |

## 5. Anomalies & Findings
Each finding rated: critical / high / medium / low.

### Finding 1 — <short title> (severity: <level>)
**What changed:** <plain-English>
**Cost impact:** <amount> over <period>
**Likely cause:** <hypothesis with evidence>
**Recommended action:** <specific>
**Owner suggestion:** <team / role>
**Estimated effort to remediate:** <hours / days>
**Estimated savings if actioned:** <amount per month>

(Repeat for each finding — usually 5-15 per period)

## 6. Optimisation Backlog
Aggregate of all actionable items, prioritised by ROI.

| # | Action | Estimated saving / mo | Effort | Owner | Status |
|---|---|---|---|---|---|
| 1 | <action> | <amount> | XS / S / M / L | <team> | Open / In progress / Done |
| 2 | ... |  |  |  |  |

## 7. Reservations / Savings Plans / Commitments
| Commitment | Provider | Coverage | Expiry | Renewal recommendation |
|---|---|---|---|---|
| EC2 Compute Savings Plan | AWS | 65% | 2027-03-15 | Re-evaluate Q1 2027 — current usage suggests downsize |
| Azure Reserved VM Instances | Azure | 80% | 2026-09-01 | Renew at 1-year, 100% upfront |
| ... |  |  |  |  |

## 8. Cost by Tag (chargeback)
Only meaningful if tag hygiene is good. Note any "untagged" bucket prominently — it's a governance flag.

| CostCentre | Application | Environment | Spend |
|---|---|---|---|
| ENG-001 | platform-api | prod | <amount> |
| ENG-001 | platform-api | dev | <amount> |
| BUS-002 | sales-portal | prod | <amount> |
| **untagged** | — | — | <amount> ← flag |

## 9. Forward Look
- **Next month:** planned changes affecting cost (<bullet list>)
- **Risk:** <events that could spike spend>
- **Opportunity:** <upcoming reservation expirations, pricing model changes>

## 10. Decisions Needed (for the review meeting)
- [ ] Approve action items 1-3 from optimisation backlog (~<saving>/mo)
- [ ] Ratify reservation renewal for <commitment>
- [ ] Acknowledge untagged spend remediation deadline
- [ ] Sign off forecast for Q+1

Example invocation

User: "/cloud-cost-reviewer — March AWS spend was £14,200 (up from £11,800 in Feb), Azure £6,400 (flat), M365 £2,900 (flat). Annual cloud budget is £180k. Top mover was a new GPU instance for ML experimentation. We just signed a 1-year EC2 Savings Plan in February."

What the skill will do:

1. Quantify: cloud is on track (£23,500 month, ~£282k forecast vs £180k budget — flag overrun risk).
2. Drill into AWS jump: GPU instance accounted for ~£2k uplift; check it's not running 24x7, recommend spot or scheduled stop.
3. Cross-check Savings Plan utilisation (since signed in Feb) — if utilisation < 80%, flag.
4. Produce 5-7 specific findings with savings estimates and owners.
5. Generate forecast revision and a "decisions needed" list for the FinOps meeting.

Notes for the requester

• CSV is fine, JSON is better. Most providers export both. Larger orgs feed the data warehouse / Looker / Power BI; for monthly review, raw exports work.
• Tags are everything. Spend you can't attribute is spend you can't optimise. If untagged is > 5% of total, it gets its own action item every month.
• "Up 20%" without context is meaningless. Always frame movers as "% but also absolute change" and tie to a cause.
• Don't confuse cost with value. A spike from a launched product driving revenue is good. A spike from a forgotten dev VM is bad. Surface both.
• Reservations / SP commitments are time-bombs. Track expiry dates. A surprise SP rollover at the wrong commitment level burns money.
• "Good" looks like: finance signs off the report without follow-up. Engineering knows what to action this week. The exec sponsor reads section 1 and trusts the trajectory.