Headstart AIPricing
Single blind audit
from £2,000
A measured error rate for outputs your model has already produced, broken down by condition rather than averaged into one number, with a bound on it and a view of where errors concentrate. About two weeks. No integration: you export a sample, we judge it blind.
Priced on sample size, not on your output volume. The sample needed for a useful bound is similar whether you produce thousands of outputs a month or millions.
Continuous monitoring
from £3,000/month
A fresh blind sample of live decisions every week, judged the same way. The error rate is tracked over time, the bound tightens as the sample grows, and you get an alert when it moves.
Priced on sampling volume.
Certification is per project by design: our own data shows judgment quality does not transfer for free across content types.
Not sure which applies? The first step is a conversation, not a purchase.
Start with a conversation →