A CFO-ready LLM cost report in minutes
As AI usage spreads across teams, finance needs a clean answer to “where is the spend going?” This generator turns a flat list of usage rows — each tagged with a team, model, use case and cost — into a structured monthly report that rolls up spend three ways: by team / cost center, by model, and by use case, each with percentage shares and visual bars.
How it works
You add one row per spend bucket: which team incurred it, which model they called, the use case, and the cost in USD for the period. The tool then aggregates those rows independently along each dimension. A support team that appears in several rows (one per model or use case) is summed into a single team total, so you see both the big picture and the detail. A copy button exports the whole thing as plain text for an email or a slide.
Why this view matters beyond the provider invoice
Most LLM provider billing dashboards show spend by API key or by project, but not by the business team consuming the output. A finance team can see that total API spend was $12,000 in a month, but cannot allocate that across Engineering, Customer Support, Marketing, and Data Science without this breakdown. The chargeback conversation — which team owns which spend — requires the cost-center view that this report produces.
Equally, the model breakdown is commercially important. Different models have very different price per token. If the marketing team is calling a frontier model for short classification tasks that a smaller, cheaper model could handle equally well, that shows up clearly in the per-model cost share.
Setting up clean source data
The report is only as useful as the tagging discipline behind it. Best practice is:
Tag at the API call level. Pass a metadata header or use separate API keys per team so provider logs carry team attribution from the start. Reconstructing team allocation from undifferentiated logs after the fact is slow and error-prone.
Standardise use-case labels. Agree a controlled vocabulary across teams before the first invoice — for example: summarisation, classification, generation, code-review, search. Free-text use-case labels from different teams will not group cleanly.
Reconcile to the invoice. The total in this report should match your actual provider bill. If it falls short, you have untagged spend sitting in a default project. Find and allocate it before distributing the report.
Tips for a clean report
- Track month over month. Run it every period and watch which cost center’s share is rising fastest — that is where an optimization review pays off.
- Flag outliers. A team whose cost doubled month-on-month without a corresponding product launch warrants a short conversation before the finance review.
- Pair with the allocator. If a team’s spend arrives as one lump sum from their API key, use the per-feature cost allocator to split it by use case before entering rows here.
- Copy into your slide deck. The plain-text output is designed to paste cleanly into a slide or email without reformatting.