How it works
From question to proof
Fresh tasks, exact marking, signed results. No AI judge.
Tell us the work your AI does: payroll, invoices, stock, claims or support.
We test it with brand-new tasks it cannot have seen. Code knows the exact answer before the model replies.
You get your score with a range, example mistakes and a signed results file.
What "fully right" means
Each task has many fields, such as every pay line, every invoice status or every claim decision. A task counts as fully right only when every field matches the answer key. Almost right counts as wrong, because in real work one wrong pay line is a wrong payslip.
Why ranges matter
A score from 120 tasks is a measurement, not a fact about the model forever. We show a 95% range next to each score, so you can see how much a result could move by chance.
What we do not do
- We do not use an AI to mark answers.
- We do not approve or endorse models or vendors. A result is a test output on a date.
- We do not need your AI provider keys for public models. See Trust and security.
Try a free AI check
Tell us what your AI does. We run fresh tasks of that type on your model and send you a short report with real mistakes, if there are any.
Get my free check