Observe
Read connected billing exports, utilisation signals, account structure, tags, budgets, ownership, and service context. Missing telemetry is labelled not collected; it is never converted into a zero-cost or zero-usage assumption.
O2 AI FinOps Agent
The cloud bill, without the fiction.
O2 AI FinOps Agent is an AI FinOps agent that reads connected AWS and Google Cloud billing and utilisation data, labels every figure with its source, forecasts spend, identifies waste, and prepares optimisations for a named owner to approve.
If a source or metric is unavailable, it stays labelled not collected. Estimates remain separate from billed, paced, and forecast totals, so zero never stands in for data the agent did not receive.
o2-finops.contextILLUSTRATIVE / SOURCE-AWARE
AWS Cost and Usage Report
Measured
Google Cloud billing export
Not connected
Utilisation telemetry
Not collected
Spend horizon
as of latest successful sync
ObservedBilled compute increased
ModelledRightsizing option prepared from measured telemetry
ControlledOwner approval required
Missing data remains “not collected”No infrastructure change before approval
Observe → Model → Propose
Cost intelligence that stops at an accountable decision.
Model
Separate measured spend from estimated scenarios, project the current run rate, and test possible demand or capacity changes. Forecasts carry their source, data age, and coverage so teams can judge the evidence before relying on it.
Propose
Prepare keep, investigate, rightsize, schedule, commitment, or storage actions with cost and workload tradeoffs attached. The agent explains the option and routes it to the named engineering or finance owner instead of silently changing infrastructure.
ACCOUNTABILITY GATE
The agent prices the tradeoff; a named owner decides; the source, reasoning, approval, and resulting action remain connected as one decision record.
One filter context
Five views. One cost truth.
Change cloud, account, period, or tags once. Dashboard, Optimization, Infrastructure, Governance, and Intelligence keep the same scope, provenance, and source freshness.
- CLOUD
- AWS / Google Cloud
- ACCOUNT
- Connected scope
- PERIOD
- Selected billing period
- TAGS
- Applied across every view
01
Dashboard
Where the money is right now
Spend you can pace, not just total.
Follow burn rate against budget, projected month-end, and a plain on-track, at-risk, or over-budget state. Ask a cost question in natural language, then inspect team allocation, workload carbon context, licence usage, and the agent's latest findings without changing the active scope.
- Budget pacing and month-end projection
- Plain-language questions over connected cost data
- Team allocation, licence, and workload context
02
Optimization
Where the money leaks
Waste with a name, an owner, and a price.
Investigate unattached storage, idle addresses, stopped compute, old snapshots, and underused Kubernetes capacity across connected clouds. Rightsizing uses measured CPU and memory; resources without those metrics are excluded and counted as not collected rather than guessed.
- Leakage and commitment coverage
- Rightsizing from measured telemetry only
- Savings path from identified to verified
03
Infrastructure
Where the money goes
Cost, broken down by the thing that incurs it.
Relate billed cost to services and resources across AWS and Google Cloud. Review expensive resources, storage class opportunities, network and egress paths, Kubernetes node pools, and managed data systems with period-over-period movement kept in context.
- Service and resource cost movement
- Storage, network, and Kubernetes context
- Managed database and warehouse cost
04
Governance
Rules that hold
Tags, budgets, and posture where teams already look.
Review required-tag coverage, budget guardrails, active exceptions, ownership, and account posture beside the resources that produced the signal. Every count carries coverage, and an expired credential is shown as a source failure instead of an empty estate.
- Tag coverage and missing ownership
- Budget policy and exception state
- Security posture with source coverage
05
Intelligence
What happens next
A forecast you can argue with.
Model sixty days of available history into a thirty-day projection with a visible confidence range. Explore demand, region, or migration scenarios, and use the configured reasoning model to prepare an anomaly investigation without allowing estimates into billed or paced totals.
- Forecast with confidence and source age
- Scenario modelling kept separate from billing
- Anomaly context routed to a named owner
Every number, with its receipt
Why the numbers hold up.
O2 AI FinOps Agent keeps measured billing, estimated scenarios, source failures, and missing telemetry visibly different. A recommendation is only as credible as the data behind it, so provenance appears with the result rather than being buried in a footnote.
evidence.contractPER RESULT
Provenance
Every figure states whether it is measured or estimated, where it came from, and when the source last succeeded.
Billed means billed
Only connected provider billing data enters billed, paced, or forecast totals. Estimates remain estimates.
Absent is not zero
A missing metric reads not collected. O2 AI FinOps Agent never fills an unknown with zero.
Failure is explicit
An empty account, an expired credential, and a source with no export are shown as different states.
Nothing simulated
No random values, demo tier, or placeholder constants enter the numbers presented as estate data.
Accountability gate
A recommendation is not a change.
O2 AI FinOps Agent can prepare an option and its expected effect. Engineering and finance owners decide whether the evidence is sufficient, when the change is safe, and which policy gate applies.
Price the tradeoff
Cost, capacity, reliability, and workload context stay together.
Name the owner
The accountable engineering or finance reviewer is explicit.
Record the outcome
Approval, exception, execution, and verification stay traceable.
Operating model comparison
From reporting cost to preparing a decision.
| Dimension | O2 AI FinOps Agent | Dashboard-based FinOps | Manual review |
|---|---|---|---|
| Signal handling | Continuously relates connected cost and usage signals. | Presents reports and filters for a person to interpret. | Starts from periodic exports and reviews. |
| Operational context | Connects spend to workload, capacity, owner, and change history. | Usually centers the billing view. | Depends on context assembled by the reviewer. |
| Action boundary | Prepares options and routes them to a named owner. | Leaves investigation and action outside the report. | Relies on tickets, meetings, and individual follow-through. |
| Decision record | Keeps source, reasoning, approval, and outcome connected. | Often requires separate notes or workflow. | Lives across spreadsheets, tickets, and messages. |
Connected operating context
Cost decisions start before the bill arrives.
Infrastructure Autodiscovery supplies inventory, dependency, and ownership context. O2 Shipper brings cost and capacity questions into the repository-to-cloud plan before production.
O2 AI FinOps Agent FAQ
Questions about source-aware cloud cost decisions.
How O2 AI FinOps Agent reads AWS and Google Cloud cost signals, separates billed data from estimates, and keeps infrastructure changes under accountable approval.
What is an AI FinOps agent?
An AI FinOps agent continuously reads connected cloud billing and utilisation data, forecasts spend, identifies waste, and prepares optimisations for a named owner to approve. O2 AI FinOps Agent labels every figure with its source, freshness, and coverage so engineering and finance teams can judge the operational tradeoff as well as the cost.
How is an AI FinOps agent different from a cost dashboard?
A cost dashboard presents reports and filters for a person to interpret. O2 AI FinOps Agent relates billing and usage changes to services, owners, capacity, and operational context, then prepares a reviewable recommendation. It keeps measured data, estimated scenarios, approval, and the resulting outcome connected instead of stopping at visibility.
Does O2 AI FinOps Agent change infrastructure automatically?
O2 AI FinOps Agent does not silently change infrastructure. It can prepare keep, investigate, rightsize, schedule, commitment, or storage options with expected cost and workload impact attached. The named engineering or finance owner reviews the evidence, decides when a change is safe, and approves, rejects, or revises it through the configured gate.
How does O2 AI FinOps Agent forecast multi-cloud spend?
O2 AI FinOps Agent reads connected AWS and Google Cloud billing and utilisation sources, separates billed cost from estimated scenarios, and models the current run rate against available workload and capacity context. Every forecast carries source freshness and coverage. Missing exports or metrics remain labelled not collected and never enter the forecast as zero.
Private beta
Start with one cloud account.
Begin with a read-only billing connection. O2 AI FinOps Agent will show source coverage, a costed inventory, and reviewable optimization candidates without changing infrastructure.