Mission control for cloud spend.
Datum is the software behind our advisory work, and you can run it too. It connects to AWS, Azure, GCP, and OCI — or takes your billing exports — then turns an AI agent loose to inventory every resource, price it against live vendor catalogs, find the waste, and put a defensible dollar figure on each fix. The engine computes the numbers. The model only explains them.
What to fix, how much it saves, and proof it worked.
Datum sits on your side of the table, takes zero provider commissions, and is cross-cloud and prescriptive: it names the resource, the action, the dollar saving, the confidence level, and the way to prove it worked. On the sample workspace we ship with, Datum inventoried 515 resources across four clouds in one run, priced them against live catalogs, and surfaced about $21,600 a month of recoverable savings on $67,010 of spend — roughly a third of the bill — split into 165 recommendations, each carrying a confidence score, an effort estimate, and a written explanation. Your own numbers come from your own data on the first run.
It finds the waste, then argues with itself.
Seven detectors sweep the estate. Every finding gets a plain-English explanation, then a second AI pass tries to knock it down before it ever reaches you. What survives is what you act on.
Seven ways to find money
The agent ingests the estate, enriches it with live pricing, and runs seven detectors in sequence — nothing is a black box.
- Idle resources, rightsizing, orphaned storage
- Storage tiering and commitment coverage gaps
- Anomaly detection and cross-cloud arbitrage
- Adversarial verification knocks down weak findings
Priced, ranked, ready to ship
Recommendations land in a hub you can triage in bulk and sort by value, then export the safe ones straight to Terraform.
- Bulk triage with confidence, effort, and dollar impact
- Approval workflows above a threshold you set
- Terraform export for recommendations as code
- Realized-savings verification closes the loop
The whole loop, not just the leak
Datum carries the full FinOps cycle, from budgets and forecasts to a month-end close, and exports in FOCUS 1.0.
- Budgets, forecasts, and what-if simulation
- Chargeback and showback that teams accept
- Month-end close with invoice reconciliation
- Commitment portfolio with utilization tracking
Six domains, one platform.
Everything a FinOps team, a platform team, and a CFO need to see the same numbers and act on them — from raw inventory to the board-ready report.
Observe
Unified dashboard, multicloud inventory with natural-language search, cost explorer, anomaly detection, a weekly AI briefing, an efficiency score, and a carbon estimate.
Autonomous agent
Seven waste detectors, a live activity feed, adversarial verification, autopilot that clears low-confidence noise, learned triage, and remediation runbooks.
Act
Recommendations hub, initiatives, approval workflows, a commitment portfolio, resource schedules, a capacity planner, and Terraform export.
FinOps discipline
Budgets, forecasts, variance narratives, chargeback and showback, month-end close, unit economics, multi-currency, and FOCUS 1.0 export.
Data in, data out
Native API connections, encrypted read-only credentials, file imports for CUR and every cloud, an AI column mapper, a REST API, signed webhooks, and share links.
Govern & collaborate
Plain-English policies compiled after each run, multi-workspace access control, audit logging, inline comments and mentions, and a command palette.
Compare the four clouds on your workload.
Before the gated platform, there is a free comparison tool anyone can use. Describe a workload or pick a preset, and it prices the same shape across AWS, Azure, GCP, and OCI, ranks the monthly totals, and shows the commitment scenarios. Every number traces to a dated, cited catalog snapshot.
Real shapes, not sticker prices
Clouds do not sell identical products, so the tool publishes an equivalence mapping and is honest about the gaps where no true equivalent exists.
- Presets: web app + Postgres, Kubernetes, LLM inference
- Ranked monthly totals across four clouds
- Commitment scenarios, on-demand vs committed
The model never generates a number.
A deterministic pricing engine computes every figure from your usage data and dated vendor catalogs. The language model writes the explanation and the summary, but it is never the source of a dollar amount, a percentage, or a rank. AI verdicts are machine-validated before they are shown. That is what makes the output something you can take to a board: every claim traces back to a source you can name.
Vendors cannot pay to alter scores, remove criticism, or change rankings. We take zero provider commissions on Datum, the same as on our advisory work.
Frequently asked questions.
Do I have to connect my cloud accounts?
Will it change anything in my accounts?
How do I know the findings are trustworthy?
We are only on one cloud. Is it still useful?
Is Datum a product or a consulting service?
Who sees the data we submit?
Run Datum on your own estate.
Access is by request during cohort onboarding. Tell us your cloud footprint and monthly spend band and we respond within one business day, usually with a pilot proposal on one month of billing data. If we cannot find the savings, you owe nothing.