Week 1
Data Ingestion: Connect and See
Week 2
Allocation: Allocate and Report
Week 3
Budgeting and Alerts: Govern
Week 4
Optimization and Enablement: Save with AI
Use AI to do the work
Each week below includes prompts you can run in the FinOps Agent or Vantage MCP to do that week’s work. Want a saved, refreshable table instead of a chat answer? Run the same prompts in Canvas. See the FinOps AI Prompt Guide for more patterns.
Use this as a baseline, not a fixed implementation order. Commercial or Enterprise teams often move at different speeds based on connected providers, access requirements, business goals, and FinOps maturity. Some teams prioritize unit costs early; others need more time on integrations, allocation, dashboards, or optimization.
Week 1: Connect and See
Days 1–7
Data Ingestion capability · primarily supports the Inform phase
Every dollar of cloud, SaaS, and AI spend is visible in one Vantage account, with at least one workspace and one shared report.
1
Connect Your Primary Providers
Connect AWS, Azure, GCP, or any combination of the three. For AWS, start at the management (payer) account so member accounts inherit. For Azure and GCP, each connection doc covers the right billing scope to start from. Add Kubernetes via the agent only after the underlying cloud provider is connected. Vantage attributes node costs back to AWS, Azure, or GCP.Then add the SaaS, data, and AI providers you already pay for: Snowflake, Databricks, Datadog, OpenAI, Anthropic, Cursor, and GitHub. Including SaaS and AI from the start, not just IaaS, is the FinOps Foundation Data Ingestion capability in practice.
2
Wait for Stable Status
Watch Settings > Integrations until each integration shows Stable. See integration status for what each state means.AWS specifically: granular filtering and grouping by resource and tag require the CUR to be ingested. Set expectations with stakeholders. The first CUR can take up to 24 hours to arrive, so those AWS views won’t have full detail at the start. See AWS Data Ingestion Delay for what’s available in the meantime.
3
Decide Your Workspace Strategy
Read workspaces and pick one of two patterns:
- One workspace per business unit: when cost data needs hard separation between groups.
- One shared workspace: when everyone should see everything.
4
Set Up RBAC If You Need Data Separation
If your organization needs hard separation between business units or regulated teams, configure RBAC now, before users start exploring. Once people have visibility into cost data, taking it back is politically harder than never granting it in the first place.Follow the RBAC Setup Guide for the recommended pattern: lock down the Everyone team, create dedicated teams scoped to specific workspaces, and assign provider integrations to the workspaces those teams should see.
Advanced RBAC (scoped team workspace access, SSO group mapping) requires an Enterprise tier account. Single-team accounts and small orgs without separation requirements can skip this step.
5
Invite Your Team
In Settings > People, click Invite People and add your FinOps stakeholders: eng leadership, finance, the executive sponsor, and the engineers who’ll act on recommendations. Assign each person an organization-level role and, if you set up RBAC above, the right team. See Invite Users to Your Account for the step-by-step flow.
6
Build Your First Cost Report
In Cost Reports, start broad: Last 30 Days, group by Service, all providers. Save it as All Spend. This is your baseline for every conversation that follows.Or with AI:
Create a Cost Report titled "All Spend" grouped by service across all providers for the last 30 days.
7
Run an "Untagged Spend" Report
Duplicate the All Spend report and filter to rows missing your single most important tag (most teams pick
team or cost-center). Save it as Untagged Spend. The goal: drive that number toward zero over the next 30 days.This pattern comes directly from how Vantage customers structure their first governance dashboard. See Cloud Cost Governance.Or with AI:Show me last month's spend that has no team tag, grouped by service.
8
Share What You've Seen
Send the All Spend and Untagged Spend reports to your FinOps stakeholders (eng leadership, finance, executive sponsor) with a one-paragraph note: “this is what we spend, this is what we can’t yet attribute, and here’s the plan to fix it.”This is the start of the shortened money-feedback loop and a direct application of the FinOps Principle that everyone takes ownership for their technology usage.
Wrapping Up Week 1
Use the following prompts along with the Definition of done below to verify the week’s work.FinOps Agent Prompts for Week 1
FinOps Agent Prompts for Week 1
Which cost providers and integrations are connected, and what's the status of each?
List my workspaces, the integrations assigned to each, and the people who have access.
List my saved Cost Reports and confirm "All Spend" and "Untagged Spend" exist.
Definition of done: all primary providers Stable, one workspace strategy chosen, RBAC configured (if applicable), stakeholders invited, two saved reports (All Spend and Untagged Spend), one summary note sent.
Week 2: Allocate and Report
Days 8–14
Allocation and Unit Economics capabilities · primarily support the Inform phase
At least 70% of cost is attributed to a known owner. The FinOps Foundation Crawl-stage sample KPI is “able to allocate at least 70% of cost to known owner.”
1
Decide Your Allocation Hierarchy Before Tagging
Pick at most three to five keys, typically
team, environment, application, cost-center, and business-unit. Document them in a shared place your engineers will actually read. The recommended flat core set and the rationale for keeping it small are in Scaling virtual tags.2
Audit Existing Provider Tag Coverage
In Settings > Tags, look at which keys your providers already emit. If
team already exists everywhere with consistent values, use it directly. Don’t invent a virtual tag rule when the underlying tag is already correct.Or with AI:What tag keys are available across my providers, and what are the most common values for the team tag?
3
Create Your First Virtual Tag
Pick Custom Value or Tag Key Collapsing to start (virtual tags, virtual tag examples). Apply the Scaling virtual tags guidance from day one:
- Scope the tag to specific providers, not All Providers.
- Set the smallest backfill window you actually report on.
- Prefer equality operators (
is,is one of) overflexible match.
Create a virtual tag called team that combines the AWS team tag and the Kubernetes organization/team label.
4
Save Filtered Views per Team
Wrap each team’s filter (provider, accounts, the
team virtual tag you just created) into a saved filter. Reports and dashboards scoped to that team can reference it, so changes to membership propagate automatically without touching every report.5
Build a Per-Team Showback Dashboard
Group your reports into folders and create a dashboard per team. This sets up showback, or visibility for each team, which the FinOps Foundation Invoicing & Chargeback capability calls “always required in any FinOps practice.” Formal chargeback to accounting cost centers is a separate question, dependent on your org’s accounting model.
6
Add Unit-Cost Context
Import a business metric and attach it to the All Spend report. Start simple (cost per GB stored, cost per token); customer-level metrics come later. This is the FinOps Foundation Unit Economics capability, and a trend on a stable denominator beats raw monthly spend.
7
Track Allocation %
Review the Untagged Spend report from Week 1 and note the percentage drop. The FinOps Maturity Model sample KPIs are “able to allocate at least 70% of cost to known owner” at Crawl, “at least 85%” at Walk, and “greater than 90%” at Run. Use your primary tag, account structure, or another documented allocation rule to measure known-owner allocation. Don’t chase Run yet. Crawl is the right target for now.
Wrapping Up Week 2
Use the following prompts along with the Definition of done below to verify the week’s work.FinOps Agent Prompts for Week 2
FinOps Agent Prompts for Week 2
Show last month's cost grouped by tag:team and call out which services still have the most untagged spend.
List my dashboards and the reports inside each.
List my business metrics and which Cost Reports they're attached to.
Definition of done: at least 70% of cost allocated to a known owner, one dashboard per team, one unit cost in place.
Week 3: Budget, Alert, and Govern
Days 15–21
Budgeting, Forecasting, and Anomaly Management capabilities · supports the Inform and Operate phases
Nothing surprises you, and every alert has an owner.
1
Connect Slack or Teams (Not Both)
Connect Slack or Microsoft Teams for notifications.
2
Add Budgets to Your Top Reports
In budgets, use a standard budget for a flat target and a hierarchical budget (up to 10 levels) when you want a parent budget broken out by team. Period format is
YYYY-MM.Or with AI:Create a monthly Budget of $50,000 for my AWS production Cost Report for the rest of the calendar year.
3
Configure Cost Alerts and Cost Anomaly Alerts
Use cost alerts for threshold and percent-change rules and cost anomaly alerts for ML-detected spikes. Send everything to one Slack or Teams channel for a week before you fan out by team. It’s much easier to tune signal-to-noise that way.Or with AI:
Create a Cost Alert that notifies my #finops Slack channel if any provider's spend exceeds 20% of last month's.
4
Schedule Report Notifications
Use report notifications for scheduled digests: daily for the engineering cost channel, weekly digest for finance, monthly summary for execs.
5
Open Issues for Your Top Three Opportunities
Use issues to track optimization work, and link each issue to the Cost Report that proves the savings. This makes optimization a tracked task with an owner, not an item on a list nobody re-reads.
6
Establish a Weekly Cadence
A 30-minute Friday review covering: anomaly alerts, budget burn, the top three issues, and the untagged % trend.
7
Spot-Check Forecasting
Open any saved Cost Report and confirm the baseline forecast is plausible one, three, and six months out. If you’ve changed filters this week, re-save the report so the forecast regenerates.
Dynamic forecast, forecasting tied to imported business metrics, is an Enterprise-tier feature.
Wrapping Up Week 3
Use the following prompts along with the Definition of done below to verify the week’s work.FinOps Agent Prompts for Week 3
FinOps Agent Prompts for Week 3
List my budgets and cost alerts, and tell me which have fired in the last 7 days.
List my report notifications and what schedule they're on.
Definition of done: budgets on the top reports, anomaly alerts wired to one channel, three issues open with owners, weekly review on the calendar.
Week 4: Optimize and Operate with AI
Days 22–30
Usage Optimization, Rate Optimization, and Education & Enablement capabilities · supports the Optimize and Operate phases
Book actual savings, put AI in the loop, and turn FinOps from a project into a practice.
1
Work the Recommendations Queue
Open cost recommendations and filter to your highest-impact provider first. Save the result as a saved view, then add it to a dashboard so progress on those recommendations stays visible alongside spend trends. This is the FinOps Foundation Usage Optimization capability in practice; for the over-provisioning patterns to look for across EC2, EBS, RDS, and container services, see Rightsizing in the Cloud Cost Handbook.Or with AI:
List my open AWS cost recommendations ranked by estimated monthly savings. Which three should I prioritize this week?
2
Audit Active Resources
Use cost recommendations to find idle and orphaned AWS resources such as unattached EBS volumes, unattached EIPs, idle EC2 instances, and idle RDS instances. Then open the related active resources or resource reports to inspect the affected resources before taking action. The cheap, no-architecture-change wins often live here.Or with AI:
List my AWS EC2 instances costing more than $100 per month and flag any that look idle.
3
Kubernetes Efficiency (If Applicable)
In Kubernetes, open the Kubernetes efficiency reports for pod-level rightsizing recommendations. See How Vantage calculates Kubernetes rightsizing recommendations for the 80% efficiency target and the formulas behind each recommendation, and Kubernetes Rightsizing in the Cloud Cost Handbook for the request/limit concepts behind it. If you don’t run Kubernetes, swap this step for another sweep through cost recommendations.
4
Financial Commitments Review
Open financial commitment reports. Financial commitment reports currently support AWS only and require the CUR to be ingested. The FinOps Maturity Model sample KPIs for commitment discount target coverage are “approximately 60%” at Crawl, “greater than 75%” at Walk, and “greater than 80%” at Run. If your commitment coverage is well below 60%, queue a commitments conversation with finance.
5
Roll Out the FinOps Agent and Vantage MCP If You Haven't Already
The FinOps Agent and Vantage MCP use the same underlying read and write tools, exposed to different audiences. Some teams will have set up one or both earlier; this week, make sure each surface is in front of the people who’d use it.
- FinOps Agent in Slack for FinOps practitioners and managers. Pick the channel where your weekly review happens. An Org Owner has to approve any AWS remediation actions (RBAC permissions). Run a few opening prompts together with a teammate:
- “Summarize this week’s spend for me.”
- “Open an issue for the largest active recommendation.”
- “Investigate the most recent anomaly.”
- Vantage MCP in the editor for engineers already in Cursor, Claude Code, or Goose. They can query Vantage from their IDE without context-switching to the console. Try: “What’s the unit cost of the checkout service this month?”
- Canvas in the console for anyone who wants the agent’s answer as a saved, refreshable table instead of a one-off chat reply. Turn the recurring prompts from earlier weeks into durable reports you can re-run. Try: “Show cost per team by service for the last three months as a table.”
6
Codify What's Stable with Terraform (If Applicable)
If your team doesn’t use Terraform, skip this step. The Vantage API covers the same resources programmatically if you’d rather script changes another way.
Write me a Terraform script for the virtual tag I just created in the Vantage console.
7
Pick Role-Based Learning Paths
Send each persona the matching Vantage University track:As a practical Vantage rollout cadence, start with one module per role per week. This supports the FinOps Foundation Education & Enablement capability.
FinOps Analyst
For practitioners owning cost visibility and allocation.
Developer
For engineers acting on recommendations.
Finance Manager
For finance partners on budgets and forecasting.
8
Look Back, Plan Forward
Run a 30-minute review with stakeholders covering allocation %, anomaly counts, issues closed and savings booked, and recommendations adopted.Then pick one FinOps capability to advance from Crawl to Walk over the next 30 days, and put it on the roadmap.
Wrapping Up Week 4
Use the following prompts along with the Definition of done below to verify the week’s work.FinOps Agent Prompts for Week 4
FinOps Agent Prompts for Week 4
What was my month-over-month change in spend, and what drove it?
What's my current commitment coverage across providers, and where could I improve?
Definition of done: at least one optimization booked, a saved recommendations view pinned to a dashboard, the FinOps Agent installed in Slack with at least one teammate using it, one engineer querying via MCP, a written “next 30 days” plan grounded in the FinOps Foundation capabilities.
You've completed the 30-day playbook
The Crawl-stage foundation is the start, not the finish. Pick one FinOps capability to advance from Crawl to Walk and rerun the playbook with a sharper focus.
Frequently Asked Questions
Do I need a paid plan to do this playbook?
Do I need a paid plan to do this playbook?
Most of it works on the free tier. The notable exceptions:
- Automated FinOps Agent remediation is for paying accounts.
- Advanced RBAC (locking down the Everyone team, scoped team workspace access) is Enterprise tier.
- Dynamic forecast tied to imported business metrics is Enterprise.
Can I move faster than 30 days?
Can I move faster than 30 days?
Yes. The pace is suggested, not strict. Many customers finish Week 1 in a couple of days. The 30-day frame is for the rhythm (weekly review, anomaly response, monthly capability advancement), not the setup. The FinOps Foundation Phases page is explicit that practitioners cycle quickly through Inform/Optimize/Operate as they mature.
We're already past Week 1. Where do we jump in?
We're already past Week 1. Where do we jump in?
Run the Definition of done check at the end of each week. Start at the first week whose checks aren’t all green. Most teams who’ve used Vantage informally for a while are missing pieces of Week 2 (allocation hierarchy, unit cost) or Week 3 (budgets and a real weekly cadence). Both are worth doing properly even if it feels like backtracking.