Skip to content
LLM Cost Audit

Pricing

One fixed price to find the money. Implementation priced on results.

No retainers, no checkout, no surprise scope. A flat assessment fee, and — if you want the fixes made — a model that ties our pay to savings you can verify yourself.

Step one

The assessment

$750one-time · non-refundable
  • Mutual NDA before any data is shared
  • Analysis of your usage export — not your prompts or billing
  • Written findings report, ranked by recoverable dollars
  • Current-vs-optimized math on every line
  • Yours to keep and implement, with or without us
Book an assessment

Step two — optional

Implementation

Flat feeor share of verified savings
  • We make the changes the report identifies
  • Priced as a flat fee or a share of verified savings
  • Results guarantee: no agreed savings, no savings-based fee
  • Savings measured as unit cost on a fixed traffic sample
  • Pay when your next bill confirms the result
Start a conversation

The honest part

We bill on unit cost, not your total bill.

Savings are measured as unit cost — cost per 1,000 calls (or per conversation, or per user) — on a fixed, agreed sample of your traffic, compared before and after.

We measure unit cost, not your total monthly bill, because your usage grows over time. This keeps the number we bill on honest and verifiable against your own invoices.

FAQ

Questions buyers ask

What exactly is included in the $750 assessment?

A structured analysis of a usage export from your OpenAI / Anthropic account, delivered as a written findings report. It identifies where spend leaks — uncached repeated context, over-powered models on low-complexity calls, async work not using the Batch API, and unmanaged context growth — and shows current-vs-optimized cost on each line, ranked by recoverable dollars. The report is yours to keep and act on, with or without implementation.

What data do you need — do I upload my prompts or billing?

No. You send a usage export from your provider dashboard after a mutual NDA is in place. You never upload prompts, customer data, or billing credentials through a web form. The assessment works from aggregate usage data — token counts, model mix, call patterns — not the content of your requests.

Is the $750 refundable?

No. The assessment is a fixed, non-refundable $750. It produces a real deliverable — a findings report with the dollar math — regardless of whether you choose to implement anything afterward.

How is implementation priced?

Implementation is separate and optional. It is either a flat fee or a share of verified savings, whichever applies to the engagement. If the agreed savings aren't achieved, the savings-based fee isn't owed.

How exactly do you measure savings?

Savings are measured as unit cost — cost per 1,000 calls, or per conversation, or per user — on a fixed, agreed sample of your traffic, compared before and after. We deliberately measure unit cost, not your total monthly bill, because your usage grows over time and a growing total would hide the per-unit improvement. Unit cost is verifiable against your own invoices, so payment can be made when your next bill confirms the result.

Why measure unit cost instead of the total monthly bill?

Because total spend grows with usage. If we billed on the total, a successful optimization could be masked by your product growing — or worse, we could appear to 'save' money in a slow month we had nothing to do with. Cost per unit isolates efficiency from growth: it's the only number that proves the optimization itself worked.

How long does the assessment take?

Typically a few business days from receiving a complete usage export, depending on the size and complexity of your workload. You'll get a clear turnaround estimate when you book.

Is my usage kept confidential?

Yes — confidentiality is the default, not an upsell. A mutual NDA is signed before any data is shared. Your identity, usage, and numbers stay private, and the client list is never disclosed to anyone.

Still deciding? See what the deliverable looks like first.