About
The independent, cross-vendor auditor for AI spend.
The independent, cross-vendor auditor for AI spend. FinOps for AI, buyer-side: it connects every dollar a company spends on AI to what its engineering teams actually ship, with a proof level on every number.
Companies now spend real money on AI — seats, tokens, and a long tail of charges that never reach a vendor console. Almost nobody can say what that spend actually bought. The tools that track SaaS spend stop at the invoice, and the vendors selling the seats have no reason to tell you which ones sit idle. Spendassay exists to close that gap.
Independence is the whole point. We don't resell AI tools, we don't take a cut of your spend, and the audit is read-only by architecture — it cannot write to the systems it reads. That's what lets us say the uncomfortable thing: which seats to cut, which renewals to renegotiate, and where the spend simply isn't working. Every figure shows its formula, so the number survives the second question in a board meeting.
How we make money
Subscription only. Nothing that moves with your bill.
Anyone claiming independence should have to say how they get paid, because that is where the incentive to shade a number lives. We charge a flat price per covered engineer per month, and that is the whole model. We take no commission on your AI spend, no referral fee from any vendor, and no share of what an audit recovers.
The last one matters most. A contingency fee would pay us more for a bigger recovery number, which is exactly the pressure that turns an estimate into a headline. Our revenue does not change based on what your audit concludes, so the number we publish is the number we found. The same commitment is in the terms of service, not just on this page.
What we refuse to do
An auditor is defined by its limits.
These five cost us features people ask for, and we hold them anyway. They are enforced in the code, not in a policy document.
- No individual rankings. We only ever report on teams of eight people or more. There is no individual view to unlock, and a team smaller than eight rolls up. This is not a policy you have to trust us on. There is nothing to turn on.
- No code or content, ever. We read seat lists, last activity, invoices and token counts — a token is the unit AI vendors bill for, roughly a few characters of text. Never your prompts, your messages or your code.
- The audit is read-only by architecture. Observe cannot merge a PR, cancel a seat or change a license. Recoveries run as step-by-step instructions your own admins carry out. Acting is Control, a separate opt-in plan (private beta) that runs on your own provider keys and can stay in shadow mode indefinitely.
- Any 'ships faster' claim comes with the bugs and rework it caused. When work ships faster, something usually gets worse: rework, bugs per PR, incidents. We show the gain next to what the speed cost you, on the same screen. A speed number without its cost is propaganda.
- We never profit from your spend. Control routes on your own provider keys, so the inference bill stays yours and never passes through us. Zero token margin, flat per engineer always, whether you route $1 or $1M. An auditor paid a share of the bill has a reason to want the bill to grow.
How the research is produced
Every figure says how sure it is.
Nothing here is published as simply true. Every number carries one of three proof levels, and the label travels with the number onto the page, into the PDF and through to the API: Counted means read from a system of record, Compared means a measured difference between matched groups, and Estimated means worked out from assumptions you can see and change — always as a range. The scale is public, as is the methodology, including the research that cuts against us.
Published vendor prices are re-checked on a 30-day cadence, and each one is stamped with the date it was last read. That check runs in our build and fails it, so a stale price stops a release rather than sitting on the price index unnoticed. Where a vendor publishes no price, we say so instead of estimating one.
Corrections
When we get something wrong, it gets logged here.
A published number that turns out to be wrong is corrected in place and recorded below, with the original claim stated rather than quietly replaced. Found one? hello@spendassay.com.
No corrections yet. This log is published empty on purpose — a corrections page that appears only once there is something awkward to put on it is one nobody ever sees.
Who runs it
Built by an operator who got the bill.
Spendassay started when its founder — an operator, not a vendor — watched an AI bill grow faster than anyone could explain it, spread across five admin consoles that each graded their own homework. The tools for tracking software spend stopped at the invoice. The productivity tools measured output and never touched the cost. Nothing joined the two, so the honest answer to a board member asking what the spend bought was a shrug, and a shrug is not an answer you can take to a renewal.
Spendassay is operator-built and independent: self-funded, no revenue from any AI vendor, and no cut of what you spend. Zero token margin, flat per engineer always. The founder works with design partners directly — every audit call is with the person who built the product, not a sales development rep.
Research and benchmarks publish under Spendassay Research, which is the company rather than a persona. The claims are meant to be checked against their sources, not against anyone's reputation.
Questions about the audit?
We answer in plain language — about the methodology, a security review, or your account.