Cloud spend was a black box. Model spend is a newer one.
Attribution, right-sizing and commitment strategy across the cloud estate, and the same discipline applied to model consumption — so a finance question and a governance question can be answered from one record instead of two.
Waste is not the interesting part. Not knowing is.
Most estates could name a number for what they spend and almost none can name what caused it. That gap is the thing worth closing, and it closes before any saving is available.
Published 18 March 2026 from 753 respondents surveyed in winter 2025, and Flexera attributes the rise to growing cost complexity from AI and new services — which is this page’s argument, made by somebody else. It is their figure, cited as theirs. Your own number is unknown until somebody looks, and it is usually distributed differently from the average.
That is the FinOps Foundation’s State of FinOps 2026, published 19 February 2026 from 1,192 respondents representing more than $83bn of annual cloud spend, and it names AI cost management the single skillset teams most need to add. Unlike a database, the thing spending the money is making its own decisions about when to. That is a control question wearing a finance costume.
Every optimization exercise that skips this step produces a list of changes nobody will approve, because nobody can say who owns the thing being changed.
Naming the reductions somebody is actually willing to make is most of the work, and it is why the assessment comes first.
One discipline, two kinds of spend.
A horizontal capability across the Reign Ops stack rather than a separate product, because the substrate that runs the workload is the thing that knows what the workload cost.
| Capability | What it means in operation |
|---|---|
| Spend attribution | Cloud and model spend attributed to the environments, teams and workloads that caused it, before anyone argues about whose it is. |
| Right-sizing | Idle and oversized resources found and fixed as a continuing practice rather than an annual project. |
| Commitment strategy | The commitment position operated rather than maintained in a spreadsheet, using insured commitments that adapt on a thirty-day cycle. How that works → |
| Model economics | Token consumption attributed by team and by workload, on the same terms as cloud. Where agents are the consumers, the identity that made the call is part of the record. The boundary that records it → |
| Cost evidence | The record of what was spent, on what, by whom, kept where it can be read rather than reconstructed at the end of a quarter. |
The insurance layer is Archera's and the terms are theirs to state. About the partnership →
With a look at the bill, not a number in a deck.
The assessment is the entry point and it is deliberately shaped so that "your position is already reasonable" is one of the outcomes.
That figure does not exist until the estate has been looked at, and quoting one before then is a sales tactic rather than an estimate.
The assessment attributes and recommends. Nothing is re-sized, re-committed or turned off during it.
Not a report with findings. Each item names the team that would have to agree to it, because that is what decides whether it happens.
Operate it with us, run it yourselves with what we found, or conclude there is not enough here to be worth doing. The assessment in full →
A cost record and a governance record are the same record, if you build it once.
The thing that knows what a model call cost is the thing that knows who made it and what policy applied. Building those as two systems is how estates end up able to answer the finance question or the control question but not both about the same event.
What we run, what you run.
Written down before anything is deployed, so the boundary is a document rather than a discovery.
| Who | What they hold |
|---|---|
| Reign Ops, automated | Operate attribution, right-sizing and the commitment position across cloud and model spend. Run the insured-commitment cycle. Produce the cost record and keep it readable. |
| Customer authored | The budget rules and the risk appetite. Which reductions are acceptable. What each environment is for. Every decision to change a commitment or turn something off. |
| Operating partner engagement | The assessment and the initial attribution. Standing up reporting your finance function will read. The periodic review as the estate changes. |
The questions that arrive every time.
How is this different from the cloud cost page?
Do you use our figures or yours?
Can you cover model spend if the models are not on your substrate?
Which deployment shapes can we have?
What does it cost?
Start with the bill you already have.
Roughly what you spend, across which clouds and which models, and whether anyone can attribute it today. The assessment is where the number comes from.